Dot com and the recent boom were because of the introduction of society changing technologies - specifically the internet, and the internet in your pocket.
Not sure where the next revolution may come from (no, 'AI' is not it).
It may be that tech will be like banking is today, still lucrative but not revolutionary/in the public eye as it was (as I understand it, banking went through it's boom in the 80s, hence Wolf of Wall Street, but I'm ready to be corrected on that).
Can you expand on this? The pace of improvement is insane. And we've barely started training on video but that is clearly imminent with big implications for robotics and entertainment.
A couple years ago I would have guessed that it would take us decades to get where we are now. The new Copilot integration into Office 365 (I can't believe I'm saying this!) looks fantastic, and even though a lot of my coworkers are skeptical, I really like using AI to help me with dev work. It doesn't do tasks for me, but it definitely helps speeds things along.
I think there's going to be a huge demand for engineers who can help integrate AI into software, regardless of the field its used in.
That seems premature. Maybe decoder transformer models aren’t it, but AI is almost absurdly broad as a field hence requires near omniscience to write off wholesale.
> Not sure where the next revolution may come from (no, 'AI' is not it).
Not sure why you say this; are you sure AI can or should be so easily dismissed out of hand without further thoughts?
---
The assertion that "AI is not it" when considering the next revolution is a point of debate. While it's true that AI hasn't yet led to a socio-economic transformation on the scale of the internet or mobile technology, it is still in the early stages of its potential impact. AI and machine learning are already influencing many industries, from healthcare with predictive diagnostics to transportation with autonomous vehicles, and their capabilities are rapidly advancing.
The part below "---" above was written by ChatGPT, to help underscore my point.
Most likely LLMs (and this is what we mean when we talk about the latest AI 'craze') will find useful purposes in some situations, almost always as an adjunct to a human (see Centaur Chess).
Perhaps LLMs' biggest contribution to social progress will be wrecking social media, as you have kindly demonstrated. Nobody wants to talk to a machine.
I don't actually know any normal (i.e. outside IT) people that have talked to or are even slightly interested in talking to ChatGPT. The user figures will have been inflated as they are with any public facing platform, the money is largely VCs jumping on the hype, and the all-time visits will be pumped by automated services attempting (and likely failing) to demonstrate value.
Weird, I do know normal people who do. I also know businesses who have it integrated with normal people talking to it. Personally, I think you're just cynical here. I'm not going to choose to take your opinion on this just because when I continue to see concrete evidence to the contrary.
… and GPT's rebuttal is basically content-free. AI has been in "the early stages" for a while now. ELIZA is more than half a century ago, and AI continues to make the basic inability-to-maintain-context level errors now that it did then. E.g., in a recent conversation with an LLM, I was consistently getting switched to the other gender by the LLM, even after attempting to correct it. In adventure stories, it'll make trivial context mistakes such as having characters whose hands are bound or otherwise incapable of motion do things that require those, e.g., pick up swords. Characters will swap or pop in or out of existence. In another example, I asked an AI recently for Kubernetes debugging advice, and after informing it that its suggestions didn't work, and asking "what would you try next?" … it repeated the same wrong suggestions.
(I am amazed that "what is the luminosity of the sun?" is finally answered not with "1 solar luminosity", which it was for a while. Now, if we can get "what is my ipv4 address" to 1-box, Goog.)
The sensor under the sink can't reliably suss out "hands" vs. "not hands", though I grant its probably a fair bit simpler in its construction … but that's sort of the point, too.
---
What I would call basic, simple problems in health tech that we don't need AI to fix:
… my own healthcare provider doesn't know who my GP with them is.
My insurance doesn't know if the bill has been paid or not.
My provider can't reliably determine that I am insured, or by who.
Providers & insurers can't agree on what the bill is, or what is being billed, and are not capable of figuring it out.
My insurer doesn't know what providers are in their network.
We still transfer health information over god-awful protocols like HL7v2 that just destroy my time as a dev supporting them.
… but yes, let's just keep trying to jam AI in somewhere, anywhere.
Automotive is the same: see the self-driving car that crushed a woman recently resulting in the suspension of the parent company's license. Car prices are through the roof, to the point where it's not economically practical to purchase at the moment, and so the end result is … I'm still driving, by hand.
They're advancing, I'll give them that, but I could feel far more meaningful impact in many places in my life from just the most basic of software engineering … or even just the most basic procedural changes (many of the problems above aren't even tech problems, but administrative ones).
I fundamentally disagree. Tech appears to still be in the middle of a decades long information technology revolution. Mobile may be complete, but it was just the latest in a series of computing revolutions. Several more waves appear to be brewing. AI/ML certainly is a candidate for another powerful wave, but so is AR/VR, wearables/ambient. All of this looks to be converging on Human-AI augmentation that appears to be both part of your mind and part of the environment in a way that will radically increase human capability. I wouldn’t be the only person who thinks this, see the Meta Smart Glasses. The first wave of this human-machine hybridization already occurred with smartphones and we are all much changed as a result.
I'll second this. We've hit peak smartphone and desktop. Next are wearables and AR. After that, maybe directly neural communication. IMO VR is still decade(s) away from being more than a novelty. AR has real practical applications that aren't hamstrung by haptic technology development.
Except I'm struggling to see what those real practical applications might be outside of a few niche fields. Even the idea of 'seeing all the pipes in the ground/wall' neglects the hard part - which is getting quality data in the first place.
As for direct neural communication, see how text replaced voice as soon as mobile devices could do it properly - nobody phones anybody nowadays. People have a tendency to take what they want from tech, and discard the rest.
> Except I'm struggling to see what those real practical applications might be outside of a few niche fields
I, for one, would be all over a permanent HUD, even just for messages and biometric data (as long as I have reasonable control over what pops up). I suspect the broader market would too if the tech gets there and the glasses don't look dorky. My point isn't that we're there yet, it's that AR is more useful and has less fundamental hurdles to wide-spread adoption than AR. I can think of many practical applications that could be made possible or enhanced with AR capabilities, but I struggle to think of any for VR that aren't just games and entertainment until we figure out strong haptic feedback systems.
What? AR is even further away than VR, because we aren’t anywhere near AR glasses tech that the average person would want to wear on a regular basis. We don’t even know how that could possibly work physically, i.e. projected opaque images on otherwise transparent glasses that could compete with the brightness of the outside environment. Not to speak of solving the vergence-accommodation conflict in such a setup.
I suspect that the law of diminishing returns comes in here - it is insanely valuable to have a map of everywhere in you pocket that you can pull out as soon as you need it. It is perhaps only slightly more valuable to have said map beamed directly into your eyes.
Regarding AI/ML - the insanely valuable step was the spreadsheet.
With the increasing pace of LLMs and the general integration of them I'm seeing in industry as well as the level of political activity around AI, and the level of innovation I see almost on the daily you really need to cite something here other than "Nyeh, I don't like it" for this to be substantive at all.
I however am looking forward to citing this comment in 5 years just like the "Dropbox is just rsync" comment.
Much of that reaction at the political levels comes from the fact that LLMs are good at the stuff Ivy League graduates are taught. The value of grammatically correct, well-referenced blithering has dropped considerably in the last two years.
There's a lot of places software can go, but currently hardware is pretty incrementally better and better. We are also getting close to the limit of how small circuits can be so the visible "runway" of tech doesn't appear to be as profitable as it once was. So now the focus of the people in the markets appears to be expanding their existing business, rather than building entire new market segments with a new killer app/product.
We are in the "boring, reliable" hardware stage where nothing is very exciting in addition to the rate hikes, so there isn't as much willingness to invest in anything risky. Pushing entrenched businesses to raise rates and enshitify is much easier to do than getting them to actually invent something new or worthwhile.
Of course, we couldn't have conceived of today's technology 100 years ago, so in 100 years we may have discovered some novel material that allows for tighter circuitry OR we'll move to biological computers or something similar.
We already have discovered such a material and it's hardly exotic. Graphite chips can be clocked a couple orders of magnitude faster than silicon. The issue is manufacturing them.
That sounds like relevant tech that could take off if there's enough investment, but isn't likely to see returns for years to come...the exact kind of thing VC seems allergic to. The bleeding edge is expensive to cut and companies are still reeling from their overly enthusiastic hiring sprees so I hope in a couple of years they will have sorted house and start paying people to play with the future.
I’m still disappointed we’re not exploring photonic circuitry or DNA computing or all the gee whiz speculative fiction post-silicon ideas from twenty years ago.
There's a guy at work who is convinced photonic circuits would solve a lot of problems. I imagine if we can start growing perfect crystals of some sort to be used in photonic chips, we can use them, but my understanding the speed of a copper circuit is already relatively close to the speed of light so that's not really the limiting factor of our current chips.
Wasn’t tech pronounced dead after the dot com bubble burst and then again after the financial crisis (rip good times). It seems to me that the reports of tech’s death greatly exaggerated.
If Microsoft hadn't been pulled back from absolute dominance in the early 00s by the justice department I think the tech ecosystem could easily have waned and become a shadow of what it ended up being.
The end of the gold rush doesn't require us to run out of ways to innovate - all it requires is that enough market power is consolidated in one place that competition from non-billionaires can be squashed like a bug. Then there's no point for anybody to innovate any more.
Remember the auto industry used to have garage-started startups too. A long time ago.
Start an innovative new tire company now and you'll either be bought for a pittance or copied by one of the big 3. They are also your main customers. It isn't a hopeful play. There aren't too many VCs looking for small auto industry startups to fund. There aren't too many auto startups in detroit.
How many things can you build for customers these days in tech that don't end up living on one of 4 megalithic platforms? It's not like the auto industry yet but it is headed in that direction.
The bursting of the dot com bubble was a good thing because a lot of stupid ideas were allowed to fail. Companies with solid business plans survived. Looking around at tech today, there are a lot of stupid ideas that need to fail ...
Funny they say that while we are literally in the middle of an AI gold rush. Before this there was crypto. After AI there will be something else. Tech is always going to attract ambitious young people looking to make it big, financially or otherwise.
If I go away and invent some cool widget tomorrow that becomes popular, I now need to start looking at additional assembly lines, parts suppliers, customs forms, and all that stuff. If I have some stupid software bullshit, like ChatGPT for cats or something, it can be as easy as putting a credit card into AWS to serve double the customers in seconds.
So long as that is the case, there'll always be pile of VC cash looking to jump into tech.
I think the claim is it's harder to win the startup lotto now than in the 2010s, which I can see being the case, with the exception of a small number of AI cos.
And at big cos it's harder to jump around as a means of increasing compensation than it used to be.
It says in 7.No Idea:
> We're going to let people apply with no idea at all. If you want, you can answer the question on the application form that asks what you're going to do with "We have no idea." If you seem really good we'll accept you anyway. We're confident we can sit down with you and cook up some promising project.
And we all know this is never gonna happen these days.
So I think the claims of this article are not valid in 2023 right? Or at least most of it, especially something like "If you don't have an idea just apply and we will give you one!”.
Greed, self-interest, and selfishness were invented by Adam Smith, duh.
In actuality I've seen internet communists regularly call clearcut examples of feudalism, mercantilism, and several other economic systems "capitalist." It often feels like they're not aware of any possible economic systems outside of capitalism and communism.
Just keep a little translation book in your head for modern speak; for example at this point "toxic" simply means "something I don't like", "literally" means figuratively, etc.
The US economy has proven this pretty decisively. There's a reason we don't have single payer healthcare, or mandatory living wage, Super PAC contribution limits, or an ability to close tax loopholes for the rich.
I worked in tech for thirty years and was able to retire comfortably. I was well paid but never had a spectacular salary - what worked was steady saving and investing.
The article says that "tech will be like banking, so the opportunities won't be as crazy as these days", but I don't know if this is true or not, even the reality tells the opposite though.
The reality is there are very few other types of businesses where you can scale your product so cheaply and still save or increase productivity to justify high prices. Sure music and entertainment can scale but at the end of the day, it's regular people buying the product and there's only so much money there.
It's a temporary issue, rates being what they are means money goes to easier profits. Just like the dotcom bust, it'll eventually make a comeback but good luck figuring out if your company will survive it.
> you can scale your product so cheaply and still save or increase productivity to justify high prices
You've just nailed an aspect of the software industry that I've never understood; if it's so easy to cheaply scale a product, then what's preventing a competitor from charging a bit less and beating your high price?
Nothing, that's why VCs will invest an enormous sum to grow faster than the competition to become the dominate player. Once established and people are used to the product other companies will struggle to compete against it without doing so purely on a cost cutting model that they can't sustain against bigger players.
Problem for many companies is that they are in the middle of a constantly changing tech landscape and become obsoleted fairly quickly or try and use their size to artificially increase margins too much and allow for competitors to eat their market.
At one point in time, a pencil was considered tech, so when pens were invented, would the same people have said "tech is dead, people aren't interested in making pencils anymore!"
Prior to the computer age, tech was aerospace, plastics. Sure the computer age brought a new large kind of tech, and that has been mined, but we are also entering a new phase where sensors are driving new capabilities in health[tech], space[tech] is no longer the domain of a few governments, transportation is being revolutionized [by tech].
With tech everywhere, nobody is going into "tech" they are going to work for companies making things, and those things either are tech or rely on tech.
Also, with the value, or lack thereof, that universities are bringing to their students, do we care what industry's those students are going into? And the article doesn't say where they are going, just where they aren't going, so didn't add value.
> Ambitious young people in search of their fortune once went into finance, then they tried tech, and now they’re looking for a new industry.
For the sake of argument, let's assume that's true, and then I have a question.
The "tech" field is full of people who went into it for the relatively easy high-paying jobs and/or the dotcom/startup lottery tickets.
The field changed dramatically in the process, from the quality of the work we do, to our goals in doing it, to how we do it, to how and for what we hire.
Is there any chance the field can suddenly improve now, due to this and other factors (e.g., let's say an interest in genuinely responsible infosec materializes someday)?
Or is slow decay from the current state more likely (as current practices remain, but less money to go around)?
AI is the next big thing, and no one really knows where that's going yet. Only AI offers the promise of large numbers of $150K - $300K soft office jobs.
There's a fair amount of big stuff that's clearly on the path to success. Solar power. Wind power. Electric cars. But that involves real work and big companies are already doing it.
There's stuff that's working well in prototype but costs too much. Self-driving cars (Waymo, not Tesla). Robotic farm machines. Takes large amounts of capital for marginal payoffs.
There's stuff that works in prototype but hasn't scaled yet. Flying cars of the big electric drone type. Solid state batteries. Major technical problems lie ahead. Slow growth.
There's natural resource stuff. Cobalt, lithium, rare earths. Hard work and boring, but profitable.
Instead of a gold rush, I'd rather build convivial technology we individuals & household can grow into and grow with. Tech as something built on high and delivered down has been the way of the past two decades, but I think that like writing computing is something that, in the long term is a powerful force for people. We need a more honest basis for man to meet machine to begin the real project, of Augmenting Intellect (Engelbart's mission).
Perhaps the next gold rush is tailored manufacturing.
For example for food. Let's imagine Catalan crème brûlée. Have sliders where you can adjust the sweetness of the curd, the darkness of the caramel, and so on.
Okay, perhaps food is not such a good example. But perhaps there are industries where people really need this and are willing to decide on a lot of details. Or perhaps for food, have a default setting for the majority.
83 comments
[ 3.1 ms ] story [ 145 ms ] threadSeems like a stretch, why would high tech salaries and another bubble not get created yet again? Already happened with dotcom and then up to now.
Not sure where the next revolution may come from (no, 'AI' is not it).
It may be that tech will be like banking is today, still lucrative but not revolutionary/in the public eye as it was (as I understand it, banking went through it's boom in the 80s, hence Wolf of Wall Street, but I'm ready to be corrected on that).
Can you expand on this? The pace of improvement is insane. And we've barely started training on video but that is clearly imminent with big implications for robotics and entertainment.
A couple years ago I would have guessed that it would take us decades to get where we are now. The new Copilot integration into Office 365 (I can't believe I'm saying this!) looks fantastic, and even though a lot of my coworkers are skeptical, I really like using AI to help me with dev work. It doesn't do tasks for me, but it definitely helps speeds things along.
I think there's going to be a huge demand for engineers who can help integrate AI into software, regardless of the field its used in.
That seems premature. Maybe decoder transformer models aren’t it, but AI is almost absurdly broad as a field hence requires near omniscience to write off wholesale.
Not sure why you say this; are you sure AI can or should be so easily dismissed out of hand without further thoughts?
---
The assertion that "AI is not it" when considering the next revolution is a point of debate. While it's true that AI hasn't yet led to a socio-economic transformation on the scale of the internet or mobile technology, it is still in the early stages of its potential impact. AI and machine learning are already influencing many industries, from healthcare with predictive diagnostics to transportation with autonomous vehicles, and their capabilities are rapidly advancing.
The part below "---" above was written by ChatGPT, to help underscore my point.
Perhaps LLMs' biggest contribution to social progress will be wrecking social media, as you have kindly demonstrated. Nobody wants to talk to a machine.
"ChatGPT has 180.5 million users, generates $80 million/month revenue, and crossed over 10 billion all-time visits (1.43B visits in August 2023)"
If you can't understand why normal people like having a human like chat interface to a machine then you're a bit out of touch.
(I am amazed that "what is the luminosity of the sun?" is finally answered not with "1 solar luminosity", which it was for a while. Now, if we can get "what is my ipv4 address" to 1-box, Goog.)
The sensor under the sink can't reliably suss out "hands" vs. "not hands", though I grant its probably a fair bit simpler in its construction … but that's sort of the point, too.
---
What I would call basic, simple problems in health tech that we don't need AI to fix:
… my own healthcare provider doesn't know who my GP with them is.
My insurance doesn't know if the bill has been paid or not.
My provider can't reliably determine that I am insured, or by who.
Providers & insurers can't agree on what the bill is, or what is being billed, and are not capable of figuring it out.
My insurer doesn't know what providers are in their network.
We still transfer health information over god-awful protocols like HL7v2 that just destroy my time as a dev supporting them.
… but yes, let's just keep trying to jam AI in somewhere, anywhere.
Automotive is the same: see the self-driving car that crushed a woman recently resulting in the suspension of the parent company's license. Car prices are through the roof, to the point where it's not economically practical to purchase at the moment, and so the end result is … I'm still driving, by hand.
They're advancing, I'll give them that, but I could feel far more meaningful impact in many places in my life from just the most basic of software engineering … or even just the most basic procedural changes (many of the problems above aren't even tech problems, but administrative ones).
As for direct neural communication, see how text replaced voice as soon as mobile devices could do it properly - nobody phones anybody nowadays. People have a tendency to take what they want from tech, and discard the rest.
I, for one, would be all over a permanent HUD, even just for messages and biometric data (as long as I have reasonable control over what pops up). I suspect the broader market would too if the tech gets there and the glasses don't look dorky. My point isn't that we're there yet, it's that AR is more useful and has less fundamental hurdles to wide-spread adoption than AR. I can think of many practical applications that could be made possible or enhanced with AR capabilities, but I struggle to think of any for VR that aren't just games and entertainment until we figure out strong haptic feedback systems.
Regarding AI/ML - the insanely valuable step was the spreadsheet.
I however am looking forward to citing this comment in 5 years just like the "Dropbox is just rsync" comment.
Much of that reaction at the political levels comes from the fact that LLMs are good at the stuff Ivy League graduates are taught. The value of grammatically correct, well-referenced blithering has dropped considerably in the last two years.
We are in the "boring, reliable" hardware stage where nothing is very exciting in addition to the rate hikes, so there isn't as much willingness to invest in anything risky. Pushing entrenched businesses to raise rates and enshitify is much easier to do than getting them to actually invent something new or worthwhile.
Of course, we couldn't have conceived of today's technology 100 years ago, so in 100 years we may have discovered some novel material that allows for tighter circuitry OR we'll move to biological computers or something similar.
The end of the gold rush doesn't require us to run out of ways to innovate - all it requires is that enough market power is consolidated in one place that competition from non-billionaires can be squashed like a bug. Then there's no point for anybody to innovate any more.
Remember the auto industry used to have garage-started startups too. A long time ago.
Start an innovative new tire company now and you'll either be bought for a pittance or copied by one of the big 3. They are also your main customers. It isn't a hopeful play. There aren't too many VCs looking for small auto industry startups to fund. There aren't too many auto startups in detroit.
How many things can you build for customers these days in tech that don't end up living on one of 4 megalithic platforms? It's not like the auto industry yet but it is headed in that direction.
there has never been this many opportunities in tech with the recent ai boom
So long as that is the case, there'll always be pile of VC cash looking to jump into tech.
And at big cos it's harder to jump around as a means of increasing compensation than it used to be.
It says in 7.No Idea: > We're going to let people apply with no idea at all. If you want, you can answer the question on the application form that asks what you're going to do with "We have no idea." If you seem really good we'll accept you anyway. We're confident we can sit down with you and cook up some promising project.
And we all know this is never gonna happen these days.
So I think the claims of this article are not valid in 2023 right? Or at least most of it, especially something like "If you don't have an idea just apply and we will give you one!”.
I'm sorry to tell you that "CatGPT" is already a crowded field with many parody entrants.
alright, it's Friday - I'll bite.
Really? Human nature to chase fortune?
In actuality I've seen internet communists regularly call clearcut examples of feudalism, mercantilism, and several other economic systems "capitalist." It often feels like they're not aware of any possible economic systems outside of capitalism and communism.
It's a temporary issue, rates being what they are means money goes to easier profits. Just like the dotcom bust, it'll eventually make a comeback but good luck figuring out if your company will survive it.
You've just nailed an aspect of the software industry that I've never understood; if it's so easy to cheaply scale a product, then what's preventing a competitor from charging a bit less and beating your high price?
Problem for many companies is that they are in the middle of a constantly changing tech landscape and become obsoleted fairly quickly or try and use their size to artificially increase margins too much and allow for competitors to eat their market.
At one point in time, a pencil was considered tech, so when pens were invented, would the same people have said "tech is dead, people aren't interested in making pencils anymore!"
Prior to the computer age, tech was aerospace, plastics. Sure the computer age brought a new large kind of tech, and that has been mined, but we are also entering a new phase where sensors are driving new capabilities in health[tech], space[tech] is no longer the domain of a few governments, transportation is being revolutionized [by tech].
With tech everywhere, nobody is going into "tech" they are going to work for companies making things, and those things either are tech or rely on tech.
Also, with the value, or lack thereof, that universities are bringing to their students, do we care what industry's those students are going into? And the article doesn't say where they are going, just where they aren't going, so didn't add value.
For the sake of argument, let's assume that's true, and then I have a question.
The "tech" field is full of people who went into it for the relatively easy high-paying jobs and/or the dotcom/startup lottery tickets.
The field changed dramatically in the process, from the quality of the work we do, to our goals in doing it, to how we do it, to how and for what we hire.
Is there any chance the field can suddenly improve now, due to this and other factors (e.g., let's say an interest in genuinely responsible infosec materializes someday)?
Or is slow decay from the current state more likely (as current practices remain, but less money to go around)?
There's a fair amount of big stuff that's clearly on the path to success. Solar power. Wind power. Electric cars. But that involves real work and big companies are already doing it.
There's stuff that's working well in prototype but costs too much. Self-driving cars (Waymo, not Tesla). Robotic farm machines. Takes large amounts of capital for marginal payoffs.
There's stuff that works in prototype but hasn't scaled yet. Flying cars of the big electric drone type. Solid state batteries. Major technical problems lie ahead. Slow growth.
There's natural resource stuff. Cobalt, lithium, rare earths. Hard work and boring, but profitable.
But if the world’s most type A people stopped coming here and distorting my local job market with 60-80 hour weeks that would be great news.
For example for food. Let's imagine Catalan crème brûlée. Have sliders where you can adjust the sweetness of the curd, the darkness of the caramel, and so on.
Okay, perhaps food is not such a good example. But perhaps there are industries where people really need this and are willing to decide on a lot of details. Or perhaps for food, have a default setting for the majority.
I don't know.