HN Explorer

Ask HN: What is the weirdest or most surreal recent technology you have seen?

264 points by O__________O ↗ HN
If possible, provide links to demos, products pages, etc.

Limit posts to one tech per post to make it easier for votes to reflect HN’s voting patterns and make it easier to filter related comments.

___

Here are some examples from CES:

https://youtube.com/watch?v=SYYXida84Nc

511 comments

[ 4.1 ms ] story [ 354 ms ] thread
kokojumbo
LTO Tape drives that run their cooling at 100 percent power when idle and have no particle filters https://news.ycombinator.com/item?id=34502596
amelius
Tape drives need to be disrupted. They are way too expensive.
RockRobotRock
Not saying this isn't true, but you made the thread and the video. Is there anyone else talking about this?
stoltzmann
This isn't that uncommon in datacenter hardware. Most of the gear does not have filters on intakes. The air should already be filtered in the facility anyways.

Running fans at 100% is slightly weirder, but also not unheard-of.

KaiserPro
most things don't have filters.

I don't think I've seen a drive failure because of it though. I think the last time the robot failed it was mostly because of the autoloader not working properly.

mkj
Tape drives (3592) in a Singapore data centre had lots of problems in the smokey burning season until they put in filtering specifically for the drives. In other office-like data transfer rooms in other cities they didn't seem to have problems, so perhaps there's something especially bad about smoke.
KaiserPro
I assume smoke has some level of tar or some other heavy sticky substance, so I could see that being a problem.

It also depends on your cooling type as well. Most of the places I worked had a closed loop system, either hot/cold aisle, or radiators on the front of the rack.

THe "zombie" rooms did pull air directly from the outside world, but they had massive filters on them. dust wasn't really a problem for us.

[deleted]
(comment deleted)
mtmail
Compost bin that requires a subscription. You send the compost via mail to a processing facility https://www.fastcompany.com/90834481/nests-co-founder-design...
klntsky
We should normalize having chickens in the office.
red-iron-pine
No. Have you ever raised chickens, or cleaned out a coop?
klntsky
FAANG companies probably have enough money for additional space and employees to look after the chickens
aj_icracked
Ha! I actually run a company that does Coop cleaning, babysitting, coop set up, day old chick delivery, etc. We're launching the world's best Coop in 45 days, stay tuned :)
adrianmonk
Although it sounds funny, this is actually kind of a serious question: are you concerned about an egg bubble?
aj_icracked
Not at all, I assume egg prices will normalize in a few months / quarters to previous consumer costs - call it $1.50-$2 a dozen. It's super odd though that we're launching a company that allows anyone with a backyard to raise chickens / get eggs. For a lot of people, eggs are a bonus byproduct of raising them. A lot of people get joy from it, reduce nearly all their food waste, fertilize their yard, teach kids on taking care of them, they'll take care of you, etc. :)
tlavoie
Curious about the "world's best coop" part. What does this mean? (Asking as someone with chickens, and some old coops now.)
aj_icracked
We're building a product at the intersection of smart home tech (think cameras with computer vision for predators, alerts, etc), light robotics (automatic doors), and services (think press a button and a person shows up to clean the coop, etc). We did this small news article a few months ago that shows what we're doing if interested https://www.kxan.com/news/local/austin/chicken-coop-sitting-...
tlavoie
Ah, cool thanks!

I think I mostly need to rebuild with better structural design, better access for me and so on. A lot of the AI applications are taken care of here by 25Kg of motivated canine intelligence.

adrianmonk
I don't think this is very practical. It would cause too many communication issues in the agile process.
klntsky
The chickens are agile enough to lay eggs on their own
ok_dad
I have chickens roaming wild outside my office window, and I assure you that meetings are not much fun with "BOK BOK BOK..." resonating constantly through the mic. They are so f-ing loud.
O__________O
It would be interesting to see an environmental impact analysis of use of mail vs a compost cycle that’s does not use mail. Find it hard to believe they’re reusing the labels, boxes, plastic wrapping, etc too.
Majestic121
I think the impact analysis should be between a compost cycle that uses mail vs no compost cycle, because I would imagine that the alternative to this is just throwing food in the trash for most people
O__________O
Possibly, though there are plenty of high-density urban areas that require separating compost materials and pick it up with trash, recycling, yard waste, etc. — so to me that would seem like a better comparison; or just compare all options.
barbazoo
Blows my mind that that's a thing. If it's urban, how is there no compost bin system. If it's rural, just have a compost bin outside. It's like people don't know how to live anymore with every part of their lives being handled by some shitty subscription service. Food, waste, what's next, a box you can ship your poop in that's 39 bucks a year?!
Majestic121
I'm not really defending the service, and I think it will fail because it looks like a logistical nightmare and I don't understand why people would pay 30 bucks for this, but if it was free it could be somewhat convenient.

Especially in an urban setting if you don't have a compost bin in your building : it's not as ubiquitous as you seem to think.

dtx1
Here in Germany we have to split our garbage and one "bucket" so to speak is compostable and gets taken by the regular garbage trucks and I assume is composted in large facilities. That seems like a more reasonable approach because it scales to a whole city doing it
throw_pm23
And then the yellow bags for plastic, which they take to SE Asia on the otherwise empty container-ships, so that it gets dumped on a beach or into a river, while Germany fills its quota for recycling.

Sorry for the cynicism, this was the case a few years ago, maybe not anymore?

super256
> this was the case a few years ago, maybe not anymore?

I'm not sure what happens with the plastic in SE Asia, but we still send most of our plastic there.

https://de.statista.com/infografik/25368/wichtigste-abnehmer...

hannob
Bullshit.

This statistic is listing the share of exported plastics that are exported to certain countries. But that's not "most of our plastic", because overall exports are around ~10%. So "most of our plastic" is either recycled or burned within Germany. (Not saying any of this is ideal, but it's not helping to spread false information.)

super256
I apologize. I chose poor wording to the point where my sentence it is becoming untrue.

I meant something like: “The majority of plastic we export to other countries is still being exported to SE Asia”.

blueflow
Majority of it is burned, as always.
martin_a
As the other comment said, it's more likely that they get burned.

I think it happened that due to waste reduction, "normal" garbage is sometimes not enough to keep the furnaces running, so they take recyclable material from the yellow bags and burn it, to keep the temperatures stable...

nicoburns
This definitely happens for plastic. But food waste does tend to actually get composted.
dtx1
I don't see what you are trying to say here? That we shouldn't even try to recycle anymore because some garbage gets shipped to SE Asia?
throw_pm23
Which is better for the environment, taking it a few kilometers outside of Berlin and burying it in a landfill that is properly insulated and managed, or shipping it across to world and dropping it on a beach in Thailand where turtles and birds get tangled into it?

The first is what happens with your regular garbage, the second is what happens (with some fraction) of the "recycled" yellow bag plastic.

I was not trying to say anything particular, just how "recycling" has become a form of gaslighting to avoid the tough question of how to produce less garbage in the first place.

coldpie
Here in Saint Paul, Minnesota, USA, they are planning to ship compostable plastic bags to homes, which you can use to put organics into the normal metal/paper/plastic recycling stream. Those bags get picked out at the recycling center and used for compost to sell to local farms/etc and (hopefully) at least pay for the program, and maybe even bring in some extra funds. No extra cost to users or hassle mailing packages of garbage (lol).
tunapizza
Same here — I live in Eastern Canada and my municipality is rolling out a program where you'll put compostables in a separate bag, but the sorting is done by the trashman.
adrianmonk
> compostable plastic bags ... normal metal/paper/plastic recycling stream

That's really clever since it avoids needing to have separate trucks.

Will they also accept leaves and yard trimmings? These are much larger in volume, so I'm not sure if they could be handled, but maybe with large bags it's fine.

> sell to local farms/etc

My city sells to farms and also sells it in bags at retail stores. It's meets standards that are supposed to make it safe to use in your backyard garden.

Also, since it costs money to put stuff in a landfill, and since this diverts from the landfill, it saves a bit of money that way too.

(Further info: https://www.austintexas.gov/department/dillo-dirt)

smallerfish
Wow - they raised $100 million. They're expecting people to pay $33/mo for the service, which includes the grinder/dehydrator bin, plus the cost of mailing the dense milled organics back to their processing facility.

Figure that actual physical expenses cost $18/mo, leaving $15/mo profit per household; on top of that, they'll have extensive marketing costs - let's say marketing & employee costs take another $10/mo, leaving $5/mo profit per household. That means they'd need 1.6 million households to subscribe for a year just to get their investors their money back, or about 1% of the potential market.

How did they possibly win over investors? If I could short this I would.

arcturus17
> How did they possibly win over investors?

Low interest rates have shifted massive amounts of capital to vc. We’re talking trillions possibly, including sovereign funds and ultrabillionaire family offices.

Are those wealth and investment managers nuts? Maybe, but I’ve concluded the most likely explanation is that they know exactly how to play the vc game: place a number of wild bets and hope that one in a hundred pays off. They sometimes seem to barely do due diligence, perhaps for efficiency and speed in allocation, and because the vc portion of their otherwise well-diversified portfolio is tiny. You and I get to partake in the schadenfreude if these stupid ventures collapse but they’re wiping their tears with the returns they make on other asset classes - stocks back then, fixed income now, etc.

If interest rates go back to the “old normal” for a sustained period of time then maybe this ends, but maybe the new normal is here to stay. We’ve been told there’d be a big cleanse of stupid or otherwise unprofitable ideas (can we really cram more SaaS solutions into every single niche under the sun?) but I’m seeing very little of the sort, other than layoffs and downsizings which are not thus far commensurate with the fireworks we were promised.

smallerfish
It's true, but the raise _just happened_, which means there's still loose VC money out there. The narrative has been that there's been a transition to more conservative funding...apparently not in this case, at least.
peyton
There’s usually a lag… raise might’ve happened >1 year ago.
michaelt
> the most likely explanation is that they know exactly how to play the vc game: place a number of wild bets and hope that one in a hundred pays off.

But isn't a key point of this theory of investing that you invest in many companies that have a 1% chance of making a 100x return - rather than investing in companies with a 0% chance?

arcturus17
Paid composting collection sounds like a ridiculous idea but if for example they manage to become a standard and get monopolistic contracts like waste management companies, on a national level… there’s your moonshot. Here, have a hundred mill.
nstart
There is the part where they expect to earn back from reselling the compost.

> Mill has partnered with two processing facilities that purchase and turn all this dried food into a chicken feed ingredient, filtering out any errant forks or inorganic materials that accidentally get tossed in the bin.

The bit about filtering out inorganic material seems a little too convenient to gloss over though.

eitland
As a kid who grew up on a small (by Norwegian standards) dairy farm I remember it confused me that food waste couldn't be sold to farms.

As a grown up, seeing all the stuff people put into food waste I can understand it.

What I still can't understand though is how many people mistake food wrapping for food..!

yourapostasy
> ...how many people mistake food wrapping for food..!

It isn't mistaking, it is not prioritizing the cognitive load for the separation effort. When I was in Japan, I was impressed with the level of fastidious compliance I witnessed in materials separation. That convinced me the talk in the US about how mainstream households could not possibly support a comprehensive cradle-to-cradle materials handling infrastructure are really talking about culture and/or necessity and not logistics.

rsync
"... I was impressed with the level of fastidious compliance ..."

Diversity is neither bad nor good - it just has benefits and costs.

You witnessed one of the costs.

On the other hand, we invented jazz and snowboarding. Those are examples of benefits.

I prefer the diversity.

SirSavary
Japan, a country famously known for lacking innovation?
rsync
Agreed.

I am a compost consumer and I have no desire to plasticize my pastures and fields the same way we’ve filled the oceans with plastic.

You can see this for yourself: just go to Home Depot and buy a bag of dirt … it’s not even micro plastics … whole objects like ballpoint pens and so on.

another2another
People used to throw their teabags in the compost, but now it seems that they've started adding a plastic mesh around some teabags (to increase the strength I guess?) so now it's basically a micro-plastics bomb.
rcme
I paid around that much to have my compost picked up from my apartment weekly. They would drop off a reuseable bucket for the upcoming week and pick up the bucket from the previous week.
Thrymr
It's a reasonable local business for a waste disposal company. How is a VC funded company expecting "tech-like" returns?
elil17
How did you figure those numbers? They don't seem related to any of the content in the article or any other content I can find about them online.
hedora
I'd pay $100-200 for the trash can thing. There is a company making a smaller one, with the idea that you toss the output into your own compost pile. It cuts down on odor, rodents and time to compost completion.

What happens to the trash can if I stop paying the subscription fee?

(Fast forward a few years, and someone will rewrite the above as a postmortem.)

twosdai
I used a local version of this type of product in a city I used to live in.

Main positive is as you mentioned, food waste smell is gone, and no rodents/animals in the trash either inside or outside.

sh1mmer
We have a Vitamix Foodcycler [1] we got on sale. It works well but the cycles take a while and it doesn’t process that much at once. Honestly we just went back to using a counter top bucket and then putting it in our municipal food waste bin.

Austin turn all that stuff into compost, and then give it away to the community (just turn up with bags) and I assume use on city run parks and so on.

[1] https://www.vitamix.com/us/en_us/shop/foodcycler-fc-50

soperj
Honestly if you have a big enough pile, just bury the waste in the compost. Or get a Green Cone Solar Digester, I've literally put hunks of meat & coconut husks and it just disappears, no rodents, no smell. I was putting buckets of compost in there every week and it would just melt into the earth.
kragen
a grinder/dehydrator for organic household garbage would be pretty useful without the service, because desiccated organic garbage won't rot and stink; you can dispose of it at your leisure thereafter, whether by composting, burning, feeding to chickens, or just dumping
gwbas1c
Makes me wonder if they're just iterating until they find a business, or trying to have some cashflow while they develop some IP.
swamp40
> How did they possibly win over investors?

Matt Rogers cofounded Nest. People will throw money at that without asking any questions. Winners often win twice.

Plus the hardware grinder/dehydrator could be used in either your own garden or picked up weekly/monthly and sold locally.

Plus it's Green and it actually makes sense to try and capture that waste. Something individuals can do to make a difference.

I agree $33/mo and a mail-back service is never going to get their $100M back.

spaceman_2020
Things like this make me realize how poor the rest of the world is, and how wealthy America is.
lm28469
Sometimes (often these days) i can't tell if these things are parodies or real products. It reminds me of Juicero: https://en.wikipedia.org/wiki/Juicero

In Europe many countries have specific trash bins for food waste, they're collected by the city and reused for compost and biogas. No subscriptions, no investors, no soon to be e-waste wifi and bluetooth connected trash can, no shipping dirt in box

jcims
AVE did a great teardown of one a couple years ago, impressive machine with a stupid purpose.
HideousKojima
In the US too, plenty of cities (even small ones) offer compost collection as part of standard trash and recycling collection.
barbazoo
Some of their assumptions seem to be very much US based where "green bins" for garden and food waste might not be a thing? I don't know but where I live in Canada, contents of the green bin definitely don't end up in the landfill like they say. So it's not even a problem that needs solving in many other places.
rubylark
In the US, it depends on the waste services provider, city, and county. I've had garbage companies that offer compost and yard waste services and some not. Sometimes it's provided through the city or county for free or for a fee, though not everywhere
huslage
We should not be mailing compost around...that sort of defeats one of the purposes. There are local composting companies/orgs in most major metropolitan areas. Mine is Compost Now.
quickthrower2
Have I got this right?

I pay $0/m, I chuck my food scraps in the garbage.

OR

I pay $33/m, I fart around with some device in my kitchen, then send them by post to some company and get nothing in return.

And from an environment point of view, it is a device that needs high energy to produce.

I actually compost. It cost say $33 one off for the bin, which sits on soil, and can take probably couple years worth of waste. I am not in an appartment, but surely the solution there is a compost on common outdoor or rooftop area, where the cost is lower because it can be shared.

jaclaz
>I pay $33/m, I fart around with some device in my kitchen, then send them by post to some company and get nothing in return.

I think you also pay for the electricity the device uses, and for replacement filters. No idea what happens in case of issues (need of repair/replacement parts/etc.), i.e. if it is included in the subscription or if it malfunctions it is on you.

TheAlchemist
This seem beyond dumb !

That being said, having a real compost in the garden is quite mind blowing. Every time I go there to throw the peels of vegetables and fruits I think - damn, almost full. Then, somehow, the worms keep eating it and the next time it's still at the same level. Nature is pretty smart after all.

Havoc
The motions on the recent Boston dynamics demo video were eerily realistic and humanlike. Sure preprogrammed but still that smoothness stood out to me as something new
code_runner
The ability for the robot to move in multiple directions at once… bending knees and arms and waist at the same time, threw me for a loop.
adlpz
I believe this is the video, for reference: https://www.youtube.com/watch?v=-e1_QhJ1EhQ
yonixw
In 0:30 you can see the bag changes (the wrinkles), could be just a reshoot, but makes the magic gone.
swamp40
The bag snaps closed internally. I had to watch it at 25% speed over and over.

The cool thing about slowing it down is you can see its feet wobbling side-to-side at about 10X per second to keep its balance.

gpanders
It's interesting to see both the advances in humanoid robotics and language models like ChatGPT. The obvious next step will be to combine these two to achieve a truly human-like android, just like in old school science fiction.

Of course, bridging the uncanny valley to make it even look passably human is another milestone entirely.

7ewis
Maybe not weird or surreal enough for HN. But I am amazed by the stability and quality of videos from my drone.

It's just amazing that I could easily fly a kilometer away whilst having a live feed of the footage being streamed back to my controller.

I have a sub 250g drone, so it's so light yet easily able to withstand pretty strong winds and the videos look great on a big 4K TV afterwards!

bilsbie
Do you worry about losing it though?
7ewis
Yeah, I have insurance with 'Flyaway Coverage'. Still have to pay a fee and there's a limit to the number of claims you can make over a time period but it's a lot cheaper than buying a new drone.
wil421
I do a lot of kayaking (not white water kayaking) and fishing off my kayak in lakes, rivers, and the ocean on calmer days.

Does your style drone have any following features? Looking for something to fly around looking for fish and something that will follow me or do fly-bys when I’m doing something with the family.

Ftuuky
Not OP but Skydio might have the drones you need?
jcims
DJI is a juggernaut and i imagine will catch up, but when i last looked about a year ago Skydio autonomous tracking was dramatically better. It was the first thing that popped into mind when reading that question.
7ewis
I have a DJI Mini 3 Pro, it has a feature called Active Track[0] where you highlight a subject on the screen and then select how you'd like it to follow you. It can go alongside you or behind you - can follow things like cars too. Despite the sensors, it can be a little scary getting it to follow you when there are trees or telephone lines nearby, but in a lake or the sea it should be great. There are safety features where if it loses connection to the controller it can hover in place, land or return to home (where it took off from).

[0] - https://store.dji.com/guides/film-like-a-pro-with-activetrac...

WrtCdEvrydy
Yeah, the DJI Mini 3 Pro is that drone... get the basic controller version.
saiya-jin
in some form, this was present on DJI drones even 2 generations ago
kerpotgh
I think Skydio has the best autonomous follow feature I’ve seen so far.
xrayarx
I thought about that too and they are all not watertight. There are Swellpros that are, but they are not hobby drones. And their specs are no match to current hobby drones.
mkurz
Which drone do you have?
7ewis
It's a DJI Mini 3 Pro. Never had any experiences with other drones, so can't compare it to others - but I love it!

The new DJI Mini 3 does look good too if you're on more of a budget.

[deleted]
(comment deleted)
googlryas
It seems like every amateur or even personal production has these big beautiful sweeping aerial shots now, when 10 years ago it would be only available to medium size productions that could rent a commercial drone, and 20 years ago only available to large productions that could afford to rent a helicopter.
tempoponet
The original "Planet Earth" was a showpiece for HD, but also these kinds of sweeping shots that people had rarely seen before. These were done with heavy cameras and complicated wire systems run through caves and forests. Today it seems almost pedestrian and quaint.
Dzugaru
Absolutely this - I was completely blown away by the technology since it was a long long childhood dream for me (to fly or control something flying remotely).

I've flown the thing off the top of the volcano (at 3000 meters), and in heavy wind and clouds and was amazed how stable it is, how easy you can fly one and the 4k video it generates is breathtaking.

One thing I still want for this is to remove 500m ceiling and being able to pierce the clouds fast - not sure if this is possible anywhere since it's probably dangerous and illegal, but that would complete my childhood dream I think (short of actually learning to fly a real small plane or a glider).

joenot443
Is now a good time to get into the hobby? I remember having an interest in college but it seemed so expensive and prohibitive. Have drones reached the same cost-value consumer sweet spot that 4K tv and e-bikes are recently enjoying?
buggy6257
Go look up the DJI Mini 2. <700 for drone plus accessory kit. 4k video with 30min battery life all flown from your controller and phone. It’s mind blowing.
benevol
Energy with magnetics: https://www.KryonEngine.org
justusthane
The great thing about perpetual motion machines is that you don't need to spend a lot of time digging through all the details, and you don't even need to be well-versed in physics, to understand that they're impossible. You just have to know that energy cannot be created or destroyed.

Two items of note from their FAQ:

> "Except for the usual forces such as mechanical friction, resistance due to air, etc., no major force opposes this movement to the point of canceling it. "

Yes, and those "usual forces" require additional energy input to overcome.

> "The correct application of the Law of Thermodynamics (a) refers to closed systems which (b) exist in a reality of which 100% is known and understood. Both (a) and (b) can only ever be true with absolute certainty in theory, but never within the limited perception of reality how humankind experiences it - in other words: How much sense does it make to think that a system which is observed in only 4 dimensions can never and will never be influenced by what is contained and happening in the following 8 (higher) dimensions?"

So they're using magnets to pull energy from other dimensions. Got it!

blackbear_
> The correct application of the Law of Thermodynamics [...] exist in a reality of which 100% is known and understood

Uh, are they really saying that thermodynamics only starts working when we have a perfect understanding of reality, and not a picosecond before that? What the actual duck.

greggsy
Perpetual motion and free energy stuff was interesting things to read about back in early 2000’s internet (around the time when conspiracy theories were kind of niche).

The only thing of interest recently has to be the EmDrive [1] - only because the rigorous testing it was put through finally put the mail in the coffin of backyard free energy nuts.

https://en.wikipedia.org/wiki/EmDrive

quickthrower2
Nuclear fission energy is for practical purposes as good as perpetual motion
Schroedingersat
Most Uranium ore is roughly as energy dense as anthracite.
pavlov
”What does Kryon mean?”

”The correct question would be: Who is Kryon? Kryon supports Humanity in many precious ways - not only by dropping hints about upcoming revolutionary technologies based on magnetism, such as (shallow depth) geothermal energy generation, (mass-scale) water desalination and magnetic engines. You are encouraged to tune in to the neverending stream of new free audio channellings.”

It seems a being from another dimension has blessed humankind with a perpetual motion machine. Where do I send my money?

notinfuriated
Does it not seem suspicious to you when you see a video of the device, and it's of a 3d model?
amir734jj
This https://youtu.be/vS0TuIPoeBs

It's just amazing

I_complete_me
It's about gecko tape, tape that imitates how a gecko's feet adhere to surfaces. By Veritasium, FYI.
incomingpain
Deep fakes certainly hit me wrong. Especially given how easy it is to do to celebrities.

It started off with mundane stuff with a great message: https://www.youtube.com/watch?v=cQ54GDm1eL0

Then South Park guys, Trey Parker, doing absolutely hilarious deep fake video. They make Al Gore and Trump say poop. https://www.youtube.com/watch?v=9WfZuNceFDM

But then basically the same time I saw deep fakes presented as true. Trump saying dumber things than normal. Biden appearing like he has late stage dementia. None of it true. As an outside observer of american politics... pretty damned concerning.

But then I read a real statistic. 95% of deep fakes... are to make fake porn. Women who have never done porn now have tons of porn made of them. Worse yet, you can't do anything about it. You dont own the copyright. There is no copyright infringement. You cant sue for royalties because they assert its not you and you never did porn.

So the USA/DHS decided they would try to police this. Try to get this all under control. They proceed to put a registered spy in charge of it who only policed 1 side of the aisle.

So this technology is destabilizing the USA and heading directly toward... https://en.wikipedia.org/wiki/Second_American_Civil_War

f0e4c2f7
Without a doubt ChatGPT. Still surprises me daily.
yreg
To piggyback, it's been almost a year and I still haven't gotten over the first time I saw DALL-E 2. Or the first time I've used GPT-2 (via AI Dungeon Google Collab sheet). Not sure when was the previous last time I've felt that sense of wonder. Perhaps the first time I've interacted with the original iPhone?

Edit: I remembered some more from the past:

- First Boston Dynamics Atlas videos.[0]

- First successful Falcon 9 landing[1].

- A video explaining how the Curiosity landing will be done and then watching the stream when it happened.[2]

- The concurrent landing of two Falcon Heavy boosters on land.[3]

[0] https://www.youtube.com/watch?v=rVlhMGQgDkY

[1] https://www.youtube.com/watch?v=ANv5UfZsvZQ

[2] https://www.youtube.com/watch?v=h2I8AoB1xgU

[3] https://www.youtube.com/watch?v=sX1Y2JMK6g8

bemmu
First time I got the "there's something special happening here" vibe was seeing the early "hallucination" images some years ago. Some examples here: https://www.businessinsider.com/these-trippy-images-show-how...

My wild guess for the next 5 years is that we'll have Stable Diffusion with the ability to create NeRF scenes.

yreg
I wasn't able to realize the scale of the shift that's about to happen back then.

---

I just remembered another example of hearing about something that seemed like pure magic:

Reconstructing what a person is seeing out of their fMRI scan (2011): https://www.youtube.com/watch?v=nsjDnYxJ0bo

O__________O
That’s most interesting aspect to me is that 100s of thousands of people were all likely aware of the developments in the field, but few actually knew what was coming down the road in the near future. It’s these types of gaps I find the most interesting when looking at technologies like this.
ctack
What is coming though?
Acutulus
I'm right there with you. Google's DeepDream I think?

That first week of DeepDream images flooding the web showed me that someone had snuck a can of worms onto the table, and while they had graciously not opened it immediately, they had maybe cracked the lid just a bit

tricky
Magnetic resonance focused high intensity ultrasound to open the blood-brain barrier in humans to deliver drugs directly to brain tissue - https://www.nature.com/articles/s41467-019-12426-9
Static2280
This is interesting to me because my mother has MS and it's _because_ the blood-brain barrier is too open that she has health issues. Surprising there's good reason to open it up.
tricky
I learned about this from a talk given by one of the authors. The guy next to me asked, "If you open up the blood-brain barrier, doesn't that mean nasty things can cross, too?" and the author was like, "yep." There are definitely risks.
monknomo
It's an interesting problem - I was reading about bipolar, which also (probably) has the issue of too open a blood-brain barrier. There is a theory that repairing the blood-brain barrier would help, but as drugs repair the barrier, it's harder to get drugs into the brain from the blood, which means the doses need to be higher, which means the side effects are worse.

I could imagine a treatment protocol of opening the blood-brain barrier and then administering a low dose of a drug, or something along those lines

HorizonXP
Hah, I worked in Kullervo's lab in the early 2010s. It is, indeed, surreal, what you can do with MR-guided focused ultrasound.
gnarcoregrizz
I'm really hopeful about focused ultrasound for non-invasive treatment of various brain diseases.
Ftuuky
Seeing a dude grow human neurons directly connected to a computer chip to make a "human AI", all from his garage [0]. It showed me what is possible with science these days if you have enough dedication and know-how. Made me feel deeply excited and scared for the incoming future.

[0] https://www.youtube.com/watch?v=V2YDApNRK3g

guerrilla
Oh yeah this one... This really disturbs me. I can't shake the feeling that it might be conscious somehow; trapped, enslaved and tortured like some kind of Black Mirror episode.
jcims
This is why I give that engineer at Google a lot of slack for his concerns that Lambda was conscious.

There’s a bit of an imperative, in my opinion, to over-index on that concern, because we could be building and torturing things for decades because people are afraid of losing their jobs.

guerrilla
Fair enough, it does make sense to have a higher tolerance for false positives than false negatives in this domain. I think that's a hard thing for people to accept though since so far it's been very obvious the objects and systems we interact with have not been conscious (except to pansychists of course.) It might be good to get practice in and build the habit early though, even if we don't think these current systems are conscious.
jcims
Exactly. In the instance it doesn't seem warranted at all, but at some point we will have to accept 'whistleblower' type accounts of the same nature with a bit less excoriation.
antipotoad
Also in a subsequent interview with the Google engineer [1], it comes out that the sentience claim was a bit of an attention grab, but with the reasonable purpose of airing the topic of AI ethics, and how machine learning is running way ahead of our thinking on its consequences.

[1]: https://www.youtube.com/watch?v=kgCUn4fQTsc

jcims
Oh wow, thank you! I hadn’t seen that one.
barbazoo
Probably not a lot. Babies when they're born have billions of neurons but believe me, there isn't much going on yet :)
guerrilla
No, I don't believe you at all. If you hurt the baby, it suffers. Inability of most people to remember those times, doesn't make that not the case.
wruza
It signals danger and performs a mechanical avoidance procedure. You react to it by feeling a protective emotion.

To know if it suffers, first you have to understand what suffering is, in scientific terms, and make sure you detach your builtin empathy from an actual process, because it’s only yours own.

guerrilla
We literally have shittons of scientific research on exactly that. Not only that but many people can remember being babies.
jaywalk
What a deeply disturbing comment.
barbazoo
What's "deeply disturbing" about that? I just meant that cognitively there isn't much going on yet.
jcims
One of the few YouTube channels that i can identify by the experiment alone.
greggsy
The thought of clandestine modding labs and neural implants brings to mind a future similar to Ghost in the Shell.
agent281
At this point I feel like people must also be trying to modify their DNA in secret. I wonder if we'll hear some horror stories in the next few years.
nuphonenuacc
Search for lactose intolerance in the linked channel. Not completely secret.
Ftuuky
Was going to say that.

He "cured"(?) his lactose intolerance with CRISPR tech (for a few months but still very impressive).

agent281
Kind of proves my point: what percentage of practitioners are content creators in most fields? How many more people might be doing this out of sight?

That said, thanks for pointing the video out. I watched the update video and it's pretty incredible.

lqet
ChatGPT. This is the only technology that truly freaked me out in my lifetime (I am 35). It competently explained to me the difference between Dijkstra's algorithm and A* in a psalm-like poem [0]. It was able to find misspelled words in a long text I just made up, list them in alphabetical order, and explain to me why I misspelled them. It correctly simulated a ChatGPT instance which was accessible via a HTTP API, and it correctly simulated a Linux machine on which this simulated remote ChatGPT-API was accessed via curl. It was able to give the responses in XML, JSON, CSV, and other formats, and it was able to simulate Linux tools like cut, grep, sort, awk, jq, sed and others. It was able to add an additional layer of a ChatGPT simulation without problems. It was able to do entity recognition in a long text I provided, and was able to list the entities in a table, with the correct wikipedia article and a summary of the article as columns. It was able to do rudimentary public transit route planning in my home town in Germany (in the form of: go the XY Station, trains will run every 15 minutes from there to the main station, from there you will have long distance lines to Z, etc.), with an apology that it does not have current schedule data for Germany and I should visit bahn.de for better results. When it failed for math questions, it was able to give me Wolfram Alpha queries to find the correct results. It was able to formulate complex SPARQL queries to query Wikidata, run these against the simulated ChatGPT instance via a curl request, grep out some results I wanted, and list them in tabular form with corresponding Wikipedia articles.

All of this in one session, as part of a pleasant conversation, basically indistinguishable from a conversation with a real human.

[0]

  > Could you explain the difference between A* and Dijkstra's algorithm in the style of a Psalm?

  < Sure, here is a Psalm-style explanation of the difference between A* and Dijkstra's algorithm:

  Hark, ye seekers of the shortest path,
  And listen well to what I say:
  There are two algorithms that ye may use,
  To guide ye on your way.

  Dijkstra's algorithm is a general one,
  That can be used in any case,
  It explores the graph from the source node,
  To find the path with the least amount of space.
  
  But A* is a more specialized tool,
  With a heuristic function to guide,
  It searches more efficiently,
  And helps ye reach the other side.
  
  So choose wisely, oh seekers of the path,
  And let these algorithms be your guide,
  For they will lead ye to your destination,
  And help ye avoid the pitfalls on the side.
jvanderbot
That's wrong. It's poetic, but incorrect. These tools are about as good as a non specialist with time to undertake some bullshitting.
paulgb
> These tools are about as good as a non specialist with time to undertake some bullshitting.

You’re not wrong, but AI being passably good at bullshitting (while mostly keeping a rhyme/meter) is still mindblowing to me.

jvanderbot
It's an amazing accomplishment technically, and that deserves to be emphasized, but it is still not producing reliable outputs.

Subjectively incredible; objectively flawed.

rpdillon
What does produce reliable outputs?
xeromal
Humans?
rpdillon
Humans, in my experience, have even more reasons to be flawed than the AI. Yes, the AI can be incorrect, but humans can be misled in what they believe the same way, and have the added problems of ego and misaligned incentives. Heck, in this very thread we have one human calling out another for being incorrect about whether or not an AI was correct. Locating sources of reliable truth is an unsolved problem AFAIK.
anigbrowl
lol no. LLMs get things wrong, but they don't lose their shit when you point it out. I find it pleasant rather than frustrating to work on problems with one; the shortcoming for me as a sort of sort of conditioned passivity that arises out of its orientation to search its own knowledge for correct answers rather than asking questions.
pxue
"confidently wrong"

That's chatgpt right now. To get it to confidently right will take years and years more

Rastonbury
I saw dallE and GPT-2 demos and though 'oh pretty cool' about a year later we've got stable diffusion that can run on phones and ChatGPT which can remember context. I thought these things would be 3-5 years out or more. The speed of development is so quick, I think confidently right LLM is around the corner

In less than a decade, I believe front line support chat support jobs and even graphic artist jobs will be made obselete

causi
In less than a decade, I believe front line support chat support jobs and even graphic artist jobs will be made obselete

We can finally get the eighty hour long book-accurate Lord of the Rings adaptation we deserve.

But seriously, I can't wait to feed a novel into one of these things and get a comic book version in return.

ilaksh
The latest models, with the right setup, can already substitute for many (not all) tasks in knowledge jobs such as graphic design, support chat, even programming. I feel like we may be looking at less than 3 or 4 years for a lot of jobs to be mostly obsolete. Or at least, the traditional version of those jobs.

It's already at the point where you start expecting any knowledge worker to be significantly more productive by leveraging these tools.

It's hard to imagine that it will be more than five years before AI tools are available that can handle almost all tasks in these types of jobs.

For example, on my website aidev.codes I just added preliminary knowledgebase support. It can reference the knowledgebase this to write code. I would say that with the code-davinci-002 model at least it seems about at the level of a junior software engineer already since it's pretty effective with close supervision by a senior peogrammer, except for the fact that it cannot interpret visual information.

Knowledgebases/embedding search can also be used right now with these models for answering support questions. The only thing holding it back from very very wide scale adoption is the problem of making up information. There are already solutions in progress for this. It's unlikely that will take more than a few years to roll out and replace the current generation of models. Google and Microsoft will probably roll out their internet-scale chat search interfaces this year even if they can't fully mitigate the hallucination problem immediately.

I would guess more like 2-3 years for many knowledge-based jobs. If you want employment/contracts you will need to be very good at leveraging AI, or people will just use the AIs instead.

browningstreet
The wall of content and interactive exchanges that will fill the internet from chatgpt generated things will become harder and harder to wade through or evade.
kevinwang
I agree that the output is often wrong despite looking right, but in this instance I don't see anything wrong in this example.
jvanderbot
Dijkstra's cannot be used as generally as described. A-star can potentially be more general, but that's highly dependent on heuristic. Dinkstras is a-star with a "min weight on shortest path first" heuristic. Because a star doesn't expand shortest paths first necessarily, it cannot necessarily find the shortest path to any node except the prescribed endpoint. Dijkstra's will find all shortest paths with cost less than the path to the endpoint, and that's helpful sometimes.
kevinwang
I actually think the robot's poetic take matches my thinking more closely than this description

Edit:

I guess in this description dijkstras is more specific, in that dijkstras is the specific instance of A* with a zero heuristic.

But I think what HAL over here was saying is that you can use dijkstras in a superset of scenarios in which you can use (non-trivial) A*, so in that sense dijkstras is more general than A*, so it's not wrong.

lqet
What is your definition of "general" here? The classic shortest path problem (which the psalm addresses) is described as finding the shortest path between two nodes, and given non-negative edge weights, Dijkstra's algorithm is perfectly general here. It will find such a shortest path in any graph, no further restrictions.

For A*, everything depends on the heuristic, as you said. If h(v) = 0, A* is equivalent to Dijkstra's algorithm, so potentially, it can be as general. But the wrong heuristic (inadmissible, inconsistent) will lead to wrong results, and so calling it a "specialized" tool is correct. The heuristic gives you a specialized version of Dijkstra's algorithm which is faster on specific graphs.

> Dinkstras is a-star with a "min weight on shortest path first" heuristic.

I am not quite sure what you mean here. Dijkstra's algorithm is A* with no heuristic. I am not sure how a "shortest path first" heuristic would look like - do you mean that Dijkstra's algorithm chooses the node to expand next based on its shortest-path cost to the target t? Even if you would construct such a heuristic h(v) = c(v, t) by explicitly calculating c(v, t) each time, this is not what Dijkstra does, as it would basically be a perfect heuristic - you would then only visit nodes on the shortest path.

If you meant that Dijkstra's algorithm is A* with a "neighboring node with shortest path from start node first" heuristic, that's also not strictly true, because expanding the nearest node first is already built-in to A*. A* does not chose the next node v with the smallest h(v), but with the smallest g(v) + h(v), where g(v) is the shortest path cost from the source node to v.

jvanderbot
> Dijkstra's algorithm is a general one, That can be used in any case,

I take issue with this statement. "any case" implies it's more general than A star, which is not true, it's a specialization of A star with (ok you win) h(v) set to nothing. It necessarily cannot solve as many problems as a star.

And Dijkstra's won't work in "any case" at all, esp graphs with negative edge weights.

The explanation struck me as glossing over too much, to the point of being misleading. Perhaps I'm being too pedantic, fine, but a more enlightened comparison would read more like our discussion than what was given.

endorphinbomber
You realize it had the constraints of a poem, and needed to take some liberty with description there?
jvanderbot
Sure sure. I'm being pedantic. But replacing two sentences would fix it.
twic
> It necessarily cannot solve as many problems as a star.

Could you give an example of a positive-weight shortest path problem with A* can solve, but Dijkstra's algorithm cannot? I do not believe there is such a thing. Or did you mean something else?

> And Dijkstra's won't work in "any case" at all, esp graphs with negative edge weights.

A* won't solve those either.

kragen
i would rather say that a* is an optimization of dijkstra's algorithm

they both solve the single-pair shortest-path problem in digraphs with nonnegative arc weights

a* which explores nodes in a different, better order thank dijkstra's algorithm, but can only be applied in cases where you can compute an admissible heuristic

that makes it less general than dijkstra's algorithm, which works to find shortest paths in any digraph with nonnegative arc weights, not just the ones where an admissible heuristic can be computed

any problem you can solve with a* can be solved with dijkstra's algorithm (usually more slowly) but the converse is not true

now, in a sense, a* with an inadmissible heuristic such as h(v)≡0 is 'more general' in the sense that it can emulate dijkstra's algorithm and also do other things; you could say that a* is a class of algorithms of which dijkstra's algorithm is one

[deleted]
(comment deleted)
JackFr
> I don't see anything wrong in this example

It's not in the style of a psalm.

kevinwang
Ah, interesting. I didn't notice that. I have noticed that chatgpt is really impressive at doing rhyming poems/songs, and many of its generated poems sound similar to each other. I also don't remember raw gpt-3 being so good at poems.

So I wonder if there was substantial fine-tuning for chatgpt specifically to reward it to generate poems, in a particular style.

And here, it's "over indexing" on that and still generating poems in that familiar style

mensetmanusman
It is passing medical exams and will be used to help diagnose issues with similar reliability as humans.
nradov
That can only be possible in a limited set of cases where the inputs necessary to make a diagnosis are all present in the patient chart. Human clinicians rely on other inputs to make most diagnoses including visually looking at the patient from multiple angles, asking them questions, manipulating and feeling the patient's body, listening (with or without stethoscope), and even smell. We are a long way from the robotics and sensor integration technology necessary to automate those steps in the diagnostic process.
tootie
We've been down the AI diagnosis path before and even a purpose-built system failed to live up. I think that GPT, much like Tesla autopilot, is going to spend a very long time ironing out it's "last mile" before anyone will trust it. The tendency for AI to be spectacularly wrong when it fails to grasp context is still pretty prevalent.
markus_zhang
Just curious is it a political issue (risky and who to sue) or a technical issue (ai has far worse judgement)?
tootie
Mostly that it faired poorly against human doctors. Real-world medical data is too messy to feed into a machine.

https://www.theregister.com/2022/01/31/machine_learning_the_...

markus_zhang
Thanks! I wonder what is the most advanced automatic technique right now. I also think it's difficult to replace nurses -- even though basic nurse work is even easier comparing to a family doctor who needs to diagnose. I figure the reason is that to replace nurse we need a lot of automatic hardware, which needs a lot of investment, unlike ChatGPT which can really replace a service agent given enough training because it doesn't involve hardware.
tootie
We have a lot. But it's really the same story that we see with software. The more sophisticated the software you grab off the shelf the more that's expected from a single worker. Modern doctor's offices are overloaded with gadgets. Most are using automatic blood pressure cuffs, digital pulse ox and all sorts of machines to test for things that used to be untestable. The need for nurses and techs to operate them all only increases.
krona
I think it captured the essential difference between the two algorithms, i.e. the heuristic, so I don't understand how this is 'bullshit'.
[deleted]
(comment deleted)
joaogui1
A* is more general than Dijkstra, as Dijkstra ia A* with the heuristic set to 0. Also it's not in the style of a psalm
vijayr02
So perfectly suited for a management role? :)
mjhay
I did read somewhere that in some tests, ChatGPT-generated MBA papers got better grades than average MBA students.
treis
They blocked people from generating performance reviews so at least some were trying.
dennis_jeeves1
Well suited for more most walks of life actually where BS reigns supreme :)
lqet
> These tools are about as good as a non specialist with time to undertake some bullshitting.

Even if you are right, you still acknowledge that ChatGPT is on the level of a non-specialist bullshitting human you can have a meaningful conversation with. This alone would be enough to freak me out. If you would've told me that this is possible 10 years ago I would've called you crazy.

jvanderbot
Subjectively incredible; objectively questionable.
throwaway675309
I know a lot of people are vividly impressed by this but to me "writing in the style of" (Victorian sonnets, poems, hip hop lyrics) is the equivalent of transfer learning and feels more like a cheap parlor trick.

While I wont deny that it has impressive summarization abilities in terms of making excellent Q&A if you're willing to vet the information, I wouldn't exactly say that chatGPT is capable of meaningful conversation. It has great powers of recollection but its capacitive powers to produce new interesting information feels highly formulaic.

hammyhavoc
I laughed when I read someone call its output "meaningful".
HideousKojima
Also psalms don't rhyme, even in the original Hebrew. Also even if it is supposed to be a poem with rhyme and meter, ChatGPT's poem doesn't have correct meter.
danielvf
Agreeing with you. Psalms typically employ parallelism not rhyming. Two examples:

  The sun will not harm you by day,
      nor the moon by night.
and

  My life is consumed by anguish
    and my years by groaning;
  my strength fails because of my affliction,[b]
    and my bones grow weak.
magarnicle
You have to tell it explicitly the correct number of syllables in a haiku, otherwise it gets it wrong.
dennis_jeeves1
'poetic': you have described it accurately. I fail to see what the hullabaloo is all about. Call me jaded.
hammyhavoc
I'll call you refreshingly grounded.
soperj
This is my exact problem with ChatGTP, it looks great when you don't know what it's talking about, but as soon as you do, it looks foolishly over-confident in it's answers which are very clearly wrong.
Keyframe
which are very clearly wrong

but not always; And that's what makes it scary. You know people will use it and trusting it.

krsw
Not much worse than a number of people I've talked with, to be honest.
twic
Ha ha only serious - for me, the most profound thing about ChatGPT and friends is that they show how much of human behaviour is not actually intelligent in some deep sense.
[deleted]
(comment deleted)
CamperBob2
Exactly. It doesn't just share our abilities, it shares our flaws, which is actually more interesting and more disturbing.

It's not a crystal ball, it's a mirror.

iliane5
The fact that a "dumb", generative model that is simply predicting the next token when given a prompt can talk so well, perform complex tasks and interact with humans in such a convincing manner is pretty fascinating.

There is obviously more to human intelligence than text based conversation but it is pretty humbling that such an aspect of ourselves can be replicated and be so convincing at such an early stage and perform better than some humans even when it makes stuff up: toddlers can't talk, kids are smarter but don't have the technical knowledge, most adults only have a few specific areas of expertise, etc.

jerry1979
I believe the data gets trained by human rankers, and it should therefore regress to their mean. That said, if they train it to seek approval, I suppose we could teach it like any other student.
anigbrowl
I don't like OpenAI's implementation defaults that much. But the thing is you can have a conversation with an LLM about bullshitting, explore the reasons why it's undesirable and inferior to sincere conversation, and then leverage those points of agreement to modify conversational behavior, at least within the scope of that conversation.
yen223
This is my main problem with HN, funnily enough.
jredwards
As long as you're armed with that information, though, it makes getting to an 80% solution (to many things) almost effortless.
CamperBob2
Which if you haven't noticed is good enough to land a seat in the United States Congress.
kragen
what's incorrect about it

as far as i can tell it seems perfectly correct but admittedly my understanding of a* search is limited to messing with amit's visualizations of it

PebblesRox
It's also not Psalm-like! The Psalms don't rhyme (even in the original Hebrew). What they do have is a parallel structure where each idea is repeated in slightly different words.

This is just generic poetry with a sprinkling of "ye's."

Der_Einzige
Not even always poetic either

Notice that it sometimes struggles to rhyme?

TuringNYC
>> That's wrong. It's poetic, but incorrect. These tools are about as good as a non specialist with time to undertake some bullshitting.

Does it really matter when it comes to real life? Kids take medical advice from TikTok influencers. Half the country believes absurd news they see on Facebook. Half the country believes anything their president says.

kevstev
Agree- so many are dismissing it because its not perfect, but this is an absolutely enormous leap from anything else I have seen in my lifetime, and this crosses the chasm IMHO between "AI is unlikely to really be able to have general applicability" to "AI is probably going to upend life as we know it in our lifetimes."

Maybe it will just peter out at this level like self driving car technology did, but I really doubt it, and it seems to me that the building blocks are already there to commercialize this and use it in a real way. The future is going to be interesting...

spaceman_2020
ChatGPT is smart enough at the first prompt, but what really stands out - freakily, I might add - is how it can remember earlier data. You can go several levels deep and it will still reference the first answer and contextualize it properly.

"Do X" is great. But "Now change X to Y in the style of Z" is what blows my mind.

kevstev
Thats interesting. When I did play with it a bit, it wasn't clear to me if it was actually doing that, but in hindsight it definitely was and was not just a coincidence. When I first used ChatGPT, I had a dog that recently passed, and she had such a big personality (a mischievous always hungry hunting dog) and was just very different than my other dog (a pitbull mix that likes to guard things), and I had thought that the two of them would be a good basis for a children's book.

I initially put in a prompt like "tell me a story about the hunting dog and the guard dog." and the results were pretty meh. I was interested to see if it was actually pulling in data from FB so I put her name in instead (she had an FB page yes). It did not seem to do much. So I put in some things like "the little hunting dog is always hungry." "the hunting dog is always looking to hunt" and then did a fuller prompt of "the hunting dog and the guard dog are going to the park. tell me a story about it" and finally got a very good fully fleshed out story about how the little dog was on the prowl for a squirrel while the guard dog watched out and it was really well written and made about as much sense as a fictional tale could. There was no awkwardness or other tells that would have made you think that this was written by anything other than a human.

Anyway, that is huge that it can do this. I do remember when I used to play with other chatbots that remembering past context was something they were unable to do, at least not well, but in most cases it seems they didn't even attempt to. ChatGPT has blown me away in many ways, but this is a very specific leap forward.

spaceman_2020
It usually always comes up with something very meh the first time around. You have to nudge it in the right direction.

I was showing it off to my wife. I asked it to write a lecture for an intro to Latin American poetry with a focus on Pablo Neruda, including a background on Chilean history.

The first result was just 5-6 paragraphs with the most rudimentary 101 on all the above topics.

Then I told it to rewrite it as if the lecture was going to be given to grad students. Next, I told it to include examples of Neruda's poetry along with some analysis. I also asked it to flesh out the segment on Chilean history, with a focus on the independence movement.

The final lecture was good enough that you could walk into a classroom of college seniors and deliver it without any edits. The first one might have been good enough for 9th graders at most.

93po
In my experience, ChatGPT was perfectly happy providing logic for a question that counteracted previous logic it provided. When I pointed out this conflict, it disagreed that it was in conflict when it clearly was.
sumtechguy
One of them I used was supposed to be 'pick-a-path' style fantasy fight the dragon sort of thing. I decided to throw the A-Team into it. It integrated them in and even knew the correct names and their context of what each char tended to do. By far not perfect. But the fact it could do even 'kind of' that. Is very interesting. 'kind of' is for some things is perfectly fine. The generated art ones are very interesting too. It reminds me a lot of early emulation discussions. Where people were arguing unless you had cycle perfect emulation nothing would work. Turns out 'kind of' close works for like 95% of the cases. Some things did need that sort of detail. But most didnt.
CamperBob2
Agree- so many are dismissing it because its not perfect

In fact, I'd submit that phenomenon as its own entry in the Weirdest or Most Surreal Tech Trend contest. People who, when confronted with a talking dog, say things like, "This talking dog is an idiot. It thinks 429 is prime, its painting looks like something from Bob Ross on acid, and the C++ code it wrote for me is full of bugs. I don't see what all the hype is about."

Some of this dismissiveness is just whistling past the graveyard, but much of it comes from people who genuinely think the human brain is something that operates by means of ethereal, ineffable, and unreproducible mystic forces.

kevstev
I feel this way about self driving cars as well. Its like even 10 years ago, the capabilities they have today would be considered mind blowing, yet they receive so much hate because they aren't yet adapted to all situations. Its baffling to me, every time I watch a video of one I get all giddy and I am like we are living in the future! Yet even from tech circles, they get so much criticism because they aren't 100% there yet.

Its just a mindset I don't understand. And on places like reddit there is definitely some astroturfing going on or at least some definite "camps" that don't like SDC technology, but on HN AFAICT its been genuine disdain. So weird.

daveguy
I think most of the criticism with self driving is around the over-hyped under-delivered "full self driving" from Musk -- it actually seems to be getting worse compared to 5 years ago.

Apparently Cruise and Waymo are doing much more impressive things, but they only get a fraction of the press.

benlivengood
I'm curious how many people who criticize full self drive have actually ridden in the car while it navigates city streets. It makes about as many mistakes as a 1-3 years of practice new driver. I don't own a Tesla so I've ridden with friends. It's a barely noticeable difference between human and FSD operation, with very few overrides necessary.
chrisco255
If they're doing more impressive things why is it not in production at scale, right now? Just earlier today there was a story of a Waymo car getting stuck in an intersection in SF. Those cars still need constant intervention and human assistance that's why.
daveguy
The miles between intervention is known for Waymo and Cruise.

Tesla refuses to publish this data.

93po
Waymo and Cruise have published their known operating areas. It's tiny and only during certain times of day (middle of the night).

Tesla operates nearly everywhere. The scale of what's happening isn't even comparable.

daveguy
Estimates from users of FSD are ~5 miles between disengagements.

https://sites.google.com/view/fsdbetacommunitytracker/home?p...

Even the best self reported areas are less than 100 miles between disengagement.

Waymo and Cruise are at around 30,000 miles between disengagement.

Tesla has released nothing.

93po
Tesla has FSD released on literally every car in the US that has purchased it. Hundreds of thousands of cars.

No one serious in the self driving space is using disengagements to pick winners or leaders on the path to commercialization. There is no standard about what qualifies as a disengagement and companies will interpret and report their disengagements differently. Those interpretations also change over time, making it even more difficult to actually understand how a company has progressed in its technology.

And again you're missing the point. Waymo/Cruise are operating in an extremely different and much more narrow context than Tesla's FSD. Let's go put Waymo on a random road in a random city and see how it fares compared to Tesla FSD. I guarantee it will be tremendously worse and no one who knows anything about the space would argue otherwise.

daveguy
Hundreds of thousands of 5 mile jaunts.

Also, why is Tesla so adamant about how many accidents occured simply with FSD disabled? Shouldn't we know how long prior to the accident FSD was engaged? I mean, hitting the brakes in panic only technically makes it "not FSD".

I'd be happy if you showed me any reported data, self or Tesla, of 30k miles before disengagement with FSD. Even within one order of magnitude instead of two.

93po
Show me data of Waymo or Cruise doing 30k miles before disengagement in the 99% of the country where they don't operate. I guarantee their stats are worse anywhere outside their carefully chosen, extremely narrow operating areas.
daveguy
If there was a single operating area that Tesla could do as well, they would be shouting it from the rooftops. They are not only silent. They are trying to argue to the state of California that they are not testing self driving capabilities past level 2. Have a great day.

Edit: one more thing ... https://m.youtube.com/watch?v=nsb2XBAIWyA

93po
Telsa has no reason to hyper focus on a few square miles in some random city and only in the middle of the night. You too.
93po
> it actually seems to be getting worse compared to 5 years ago.

You pretty clearly don't know what you're talking about.

> Apparently Cruise and Waymo are doing much more impressive things, but they only get a fraction of the press.

They are doing objectively much less impressive things by relying on sensors that require much less intelligent analysis and will lead them to dead ends at the higher end of abilities that Tesla's approach won't.

[deleted]
(comment deleted)
CamperBob2
Agreed. However, in the detractors' defense, self-driving cars were hyped beyond all reason by people like Musk, so it's easy to justify some very vocal pushback.

GPT wasn't hyped much at all, in comparison. It just appeared. The next couple of iterations will be interesting to watch, since right now we don't really know what the rate of improvement is going to be. We might be fighting Skynet next year, or we might be in for another 50-year-long AI winter.

muzani
I consider it more like the golden calf that people worshipped while Moses was away. The Quranic version is it was made of gold, it made a sound, and so people idolized it.

I think it's a good parable. It's something expensive and man made. Even though it's on average dumber than a typical human, people try to ascribe godliness to it. And so many judge whether it has achieved omniscience. For many of us, it's just a better tool, an IDE, but for some, it's an oracle or divination tool that they use to decide whether they should marry someone.

I think the term "artificial intelligence" makes it even more confusing.

ozim
Many are dismissing it because it is just neural network. Which were basically at academic works since 80’s or earlier. It is only that now we have enough hardware and material for training neural nets that they start to be useful.
btilly
There is a lot more than that.

1. We've learned that relu is a really good function to use for neural networks.

2. We've learned that regularization is essential to having large networks without overfitting.

3. We've learned how to scale models up and down so that you can train something on a giant cluster then run it on your phone.

4. We've learned how to apply this to lots of specific kinds of problems.

etc, etc, etc.

We do NOT simply do what we did in the 80s then throw more hardware at it. Instead we are getting better at making the hardware do something useful.

ActorNightly
>nd this crosses the chasm IMHO between "AI is unlikely to really be able to have general applicability" to "AI is probably going to upend life as we know it in our lifetimes."

ChatGPT3 isn't really the definitive sign of the latter. Its just a stepping stone on the way. As soon as humans discovered differential calculus, that was the genesis to the eventual development of a general AI. Everything else along the way is an issue of optimization and spending raw compute time, whether its done by humans or machines.

jredwards
I had the same reaction. It's the first time in my life where I've ever thought, "This is the beginning of an era. They'll give a name to this period of time." And yeah, maybe that's overselling it, but I had that thought.

I've seen so many pieces of incredible technology struggle to find a use-case. By contrast, ChatGPT almost has too many use-cases to catalog, and while it's far from perfect, it seems obviously, to me, to be a tipping point of sorts.

Gordonjcp
Which metre is that supposed to fit? I can't fit that around any psalm I can think of, most of which are 8/6/8/6 metre.
jvanderbot
Shhhh. Just more bullshitting from the computer bullshit machine.
jonnycomputer
A few months ago, I would have said Stable Diffusion, but ChatGPT wins, hands down.

We're living in exponential times.

WastingMyTime89
Yet it will happily tell you random non sense with utter certainty like that it takes twenty minutes to boil a whole chicken in the middle of a recipe or listing an incorrect number of children with fake names while writing a biography.
nemo44x
I agree, it’s amazing and it’s just the beginning. When this get numerous iterations of improvement and is distributed at scale and integrated with everything it will power the next technological leap forward. This to me is obvious.
quanto
could you provide some examples of your prompts leading to simulation of Linux within ChatGPT?
Acutulus
It's an incredibly striking piece of technology and really shows which way the future winds will likely be blowing. A particular exchange I had with it left a mark on me, in which I requested it to behave like a used car salesman attempting to close a deal with a hungry customer. It proceeded to repurpose idioms, make dad-jokes and sling double entendres about the shared joy of cars and hamburgers and how they both are simple, dependable and iconic in my way of life. I sat there for a minute reading over it, nearly in disbelief.

I can accept a system such as ChatGPT synthesizing from data it sucked in, making educated guesses and so on. But to see it do such lingustic gymnastics with a very non-concrete request was humbling. It's given me a lot of pause about the way I absorb digital information and the varying degrees to which I have implicitly assumed the reliability of that information; site A slightly more trustworthy than site B, et cetera. To me, that old trust heuristic I relied on, one I have honed thanks to unfettered broadband for two decades, is now completely upended. It probably has been for some time to be fair, but my time with ChatGPT really cemented that feeling.

Every freshly written statement that comes to me through an internet connected device now gets a side-eye by default.

alex_duf
>I can accept a system such as ChatGPT synthesizing from data it sucked in, making educated guesses and so on

On that note I find it interesting that this has sometimes been an argument to dismiss chatGPT as "non intelligent". What are we if not statistical machines, synthesizing from the data we've sucked in over our lifetime?

WoodenChair
> What are we if not statistical machines, synthesizing from the data we've sucked in over our lifetime?

Reasoning machines who are actually pretty bad at statistics.

yakubin
Reasoning and statistical machines who are actually pretty bad at both.
swagasaurus-rex
We are heuristical which may not be logical, but it makes sense from an organism perspective.

For example since the world is uncertain and chaotic we seek either to conserve or expend for future gains. This can explain many stock trading behavior, despite often being a non optimal viewpoint

Acutulus
I certainly can't see us humans as anything other than that. But if that's true, us human machines seem to have a large number of "low level programs" running in the background that serve to blur the boundary between us and the machines we build and help us to elevate ourselves to a special typing in the universe that very well may not exist.

I feel I think too much about mundane things and sometimes about wondrously confusing things often to my detriment. And at least a few times a week I would swear I run into a person with the opposite problem. Their behavior, choices, preferences. Their thought processes (at least what they share with me) come off as closer to machine than human.

Something in my gut tells me we are more similar to some of the things we build than we might care to admit.

raptor556
I've been talking about this for awhile now, but I used to run a marketing service that streamed all reddit content in real time and did text analysis and bot detection. It's definitely a rough estimate but about ~65% of text content was determined to be a bot. I am entirely convinced that there are large entities (political campaigns, nations, etc.) that are using bot networks on social media sites like reddit to simulate "consensus" in online discussions and thus gently sway public opinion.
TeMPOraL
> It's definitely a rough estimate but about ~65% of text content was determined to be a bot.

A scary number. I wonder about a per-subreddit distribution, though. I imagine the primary subreddits have slightly worse human-to-bot ratio, niche subreddit somewhat better, with non-political, non-easily-monetizable subreddits having the best.

Did your analysis also attempt to identify troll farms? Would the content produced by protein bots be grouped in the ~65% of bot content, or the remaining 35%?

a123b456c
Ok how do we know you're not a bot?
friendlyHornet
Dear fellow human,

you have won cake. There will be a party in your honor

Please lie on your stomach with your arms at your sides. A party associate will arrive shortly to collect you for your party.

SqueamOssi
Yeah - I can't believe how much this HN discussion seems like a human generated thread.
anigbrowl
Spot on. I am interested in any publications that might have come out of your research if you care to share.
HeyLaughingBoy
> large entities (political campaigns, nations, etc.) that are using bot networks on social media sites

I think this has been conventional wisdom for a few years now.

93po
It's wild how low quality so many of the comments are on reddit, to the point that it makes me wonder "Why did this person comment something so empty and non-contributing to a post that already had 3000 comments?"

I don't know whether to believe people are so wasteful of their own time or whether this is just low-effort bot posting to build consensus. Combined with how harshly and instantly main subreddits like /r/politics and /r/news shadow ban accounts, it's basically impossible to dissent

JackFr
That's not even close to the style of a Psalm. It's still pretty cool though.
agalunar
Ought to be "to guide you on your way", "helps you reach the other side", "for they will lead you", "and help you", since "ye" is the plural nominative (not accusative).

I grew up using archaic, Jamesian English regularly (in a religious community), so it always sounds wrong when people use them incorrectly while attempting to be old-timey, which is a bit distracting and ruins the effect.

Apparently ChatGPT suffers from hypercorrection as much as any modern speaker.

(The meter is also dreadful, which is another thing that I think many modern English speakers don't often do precisely.)

rsync
RE: chatGPT ...

There was an "ask HN" a week or so ago about knowing how your personal machine was or was not compromised ...

I am imagining training a model on all of the PDFs and email and receipts and contracts that I have spent the last 30 years protecting ... and then exploding my home directory with 2M parallel copies.

Now what ?

Which is the real correspondence ? Which contracts are real ? Which invoices and receipts ?

Could one construct a framework where all data at rest is worthless and only witnessed transactions and traffic have any surveillance value ?

dredmorbius
Chaffing on steroids.

On the other hand, note that you'd also need some way of distinguishing ... wheat from chaff ... for your own ability to track your correspondence, transactions, records, etc.

Otherwise, that is an interesting idea, particularly in a world in which storage is cheap relative to the size of documents. We're rapidly approaching an era in which storage capacities sufficient to hold the entire Library of Congress collection of books (as text) on a residential PC or server will be highly viable.

The larger problem is probably that your copy of that data isn't particularly important in most cases, it's what lies on others servers that's been generated by you / snarfed from your devices / activities. And that's far harder to chaff.

Another issue would be that any generated content that raises / matches suspicions could, in a regime with poor civil liberties protections, simply become more ammunition used against you. The old Cardinal Richelieu "six lines" dictum, though in this case, not necessarily in your hand.

wprl
I suppose this is nit picking, however that is not in the style of a Psalm.
wnolens
Until ChatGPT, the term "artificial intelligence" felt like a misnomer for any of the tech that came out in the last 10y (which seemed mostly like statistical techniques for solving specific optimization problems).
skyde
For people familiar with how ChatGpT work, is there a youtube video or blog you recommend that explain how it work.

I have a very good understanding of how deeplearning neural network work and language model like "Word2vec" that convert a sentence or a word to a vector space.

But I completely fail to understand how Given an initial text as prompt, ChatGpt will produce text that continues the prompt.

skyde
I cant believe it simply use self-attention to generate one word at a time. It got to be more complex than that.
MiddleEndian
I've spent far too much time antagonizing ChatGPT, it's good shit.
alamortsubite
This video on Helion impressed me, though admittedly I know very little about nuclear fusion: https://www.youtube.com/watch?v=_bDXXWQxK38
O__________O
Worth nothing Helion has raised $577.8M of which $375M was provided by Sam Altman.
terramex
Youtube recommended me a good rebuttal of that video yesterday: https://www.youtube.com/watch?v=3vUPhsFoniw
O__________O
Would be super curious to hear Sam’s response to that video given it claims that it’s impossible for it to be a clean reaction, which contradicts statements I believe have heard Sam make about it.
agumonkey
not sure if it qualifies as technology:

- nicolas schabanel team works on molecular DNA programming, they had game of life capable structures made of DNA tiles. The mentioned being capable of transpiling tiny lambda calculus expressions but were stuck (this was in 2013~) due to thermal degradation on long programs. His courses are on dailymotion.

- michael levin morphogenesis / developmental biology research: dude revived 50s~ ideas with gap junctions and cell levels "information processing" .. he ended up regrowing a second head on a larva after reprogramming cell messaging to alter the natural regeneration process. He was in a lot of interviews/podcasts in the last months.

jcims
I truly feel there’s going to be some kind of industrial revolution with DNA or derivative based manufacturing.
agumonkey
A lot of trends are showing nanotech reaching mainstream. Just two days ago there was a veritasium video about gecko skin like thin film that could hold smooth objects passively.
v8xi
Its just getting revved up :)
leesec
Seconding Michael Levin, that guy is doing some incredible stuff
agumonkey
It made me feel something very very new, a blend of infinite "power" at reach and at the same time something that could really break sanity and ethics forever.
HideousKojima
It definitely comes straight out of sci-fi at least. In the original Deus Ex, I think when you first break into the Versalife offices in Hong Kong, you can overhear an employee talking about their universal constructors (basically 3d printers that work at the atomic level). The employee mentions one of his friends created Conway's Game of Life with buckyballs.

Edit: nevermind it's a Versalife employee at a nearby bar/club, not in the offices themselves.

smrtinsert
AI nsfw bf/gf apps. It probably just makes sense to this generation but for me, it feels like something out 90s era sci fi.
Kelteseth
This always reminds me of the film "Her"

https://www.imdb.com/title/tt1798709/

yamtaddle
And Blade Runner 2049.

And Star Trek TNG, with the holodeck (LOL, friggin' Riker)

A bunch of other stuff if ones with robotic bodies count. Star Trek: TOS has some. Fritz Lang's Metropolis (1927) features a robot re-creation of a character's dead lover. Though that's a bit different than one simulated without a reference.

sergiotapia
ChatGPT. I'm 33 and first time I've ever truly been blown away.

It's been really valuable as well not just a novelty. I'm certain it will change almost every aspect of our lives, starting with our children and how they learn in our current education pipelines. The bubble don' popped!

omginternets
I remember hearing about an GAN-based tool that would transform clothed pictures of women into nudes. That was depressingly surreal.
Llamamoe
The fact that somebody making this feels like an inevitability is just incredibly depressing. I mean, sure, it's just a fake AI guess, but Jesus Christ come on :-/
wruza
Say that you’ve never used your imagination and took no interest in this. Much easier to say come on when your hormone levels plunge back to sanity.
simonwingtips
And oddly enough, there's also a new browser plug-in that uses AI to auto-censor images client-side. https://pury.fi/
omginternets
I find this just as baffling, tbh. There are two kinds of excesses: anorexia and compulsive overeating...
jeffwask
The Mobility Scooter craze. South Park nailed it. One day there were a couple scooters and you went oh interesting. A moth later there were thousands clogging every sidewalk with no discernible method of managing these devices. Then just like they came they were gone.

How many billions spent? How much e-waste was generated?

lopis
I think we're well past the craze curve. I see way less scooters nowadays. Somewhere there is a mile high mountain of old scooters.
FeistySkink
Unfortunately it's still a thing almost everywhere in EU.
LeonM
Unfortunately, the EU country I live in does not allow them.

And that is unfortunate, because whenever I visit another EU country I find them very useful.

alistairSH
I doubt many people would argue against their usefulness. The problem is management of the scooters when not in use.

Even out in the suburbs, the local sidewalks have them discarded haphazardly. Not in a charging stand, not in a group, just ad hoc lying across the sidewalk. The bike share has dedicated stands for groups of 10-20 bikes - this makes far more sense to me than users just dropping them in the sidewalk or street when they're done riding.

If you're an older pedestrian, or otherwise mobility challenged (but able to walk), they're a nightmare.

ericpauley
This feels like largely an issue of societal norms. People leave their cars sitting all over the place and nobody complains.

For sure there are places they shouldn’t be left, like right in the middle of the sidewalk. But maybe the expectation of every scooter being in a dock is too extreme.

alistairSH
People don't generally park their cars in yards, across sidewalks, or in the woods. All places I've come across electric push scooters. Sure, cars are ubiquitous, but people mostly park them in designated parking spaces (also ubiquitous) or risk a parking ticket, boot, or tow.

DC adopted a parking zone strategy for the scooters. Haven't been downtown in a while, but imagine that could work without dedicated docks.

DangitBobby
IDK, it's pretty much a social norm to put away your shopping cart, and yet...

People are selfish and will do anti social shit for convenience if they can get away with it.

Cars only go where they belong because the state can and will punish you with towing or fines, making the reward of being selfish and anti social not worth the risk.

unsupp0rted
It's moot to me because I find the 0.0001% of space around me taken up by electric scooters vs the 99.9999% of space around me taken up by cars to be more egregious.
jollyllama
Mobility scooters are the things disabled people ride. There should be a term to refer to these electric, Razor-style form factor scooters, but I think they are just pushing the Vespa-style scooters (many of which are also electric) into the term moped and taking the term "scooter" for themselves. I call them kid scooters.
[deleted]
(comment deleted)
pavon
I've always seen the terms kick-scooter and motor-scooter to disambiguate the later two.
ausbah
they are still very prevalent in seattle. not as much of an issue so long as ppl stick to empty sidewalks and bike lanes
kirse
IMO central-city cores need to get redesigned around Bio/EV-only transport like walking, bikes, scooters, etc. You drive up to various parking-only rings where there is boatloads of parking and EV-transport docking stations and then walking/bikes/EVs get you last-mile accessibility to buildings, shops, etc.

Scooters do annoy me at times as a pedestrian but its mainly b/c city infra wasn't really designed for this middle ground, as cyclists can attest. Plus I am equally as hypocritical for sometimes zipping around on a scooter on sidewalks, so I don't complain too much. If I ever get a shot at building up a town/city this is one of my reqs.

jonnycomputer
I mean, is there really any question? There is only one answer. Okay, maybe two. Stable Diffusion and ChatGPT.

Lot's of surreal stuff out there, but these two ... there is no question. Not one.

Living in exponential times.

mrkarezina
I was thinking the same thing until I saw things like this in the thread: https://news.ycombinator.com/item?id=34503617

Stable Diffusion and ChatGPT are cool, but once you get a better understanding of the model and how it's trained it doesn't seem that exciting personally.

But being able to engineer living organisms? That's sounds like something out of blade runner.

Thebroser
Spatial Transcriptomics. Being able to see gene expression at a resolution of a handful of cells in a cross section of tissue is revolutionizing disease research. https://www.nature.com/articles/s41592-020-01033-y
JZL003
Cool I work in spatial transcriptomics, it's still early days but not too far off we're going to get great resolution and great capture. Especially for the brain, and maybe even tumor micro-environments it could be really cool
yread
All this spatial stuff is mind blowing: CODEX, CyTOF, ...
entropicgravity
CRISPR gene editing. Rewrite every instance of your genome in situ. And you not only survive but thrive, unbelievable.