Really frustrating that our privacy eroded to the point anyone would think this is reasonable.
But it does explain some conversations with Gemini when I wonder why it's talking about particular things. Like asking it about consultation processes for changes to parks only to get a response that's uncannily tuned to the email I got prompting the enquiry.
There are 5 major frontier labs competing in the US alone with serious capital. Perhaps only 2 of them are are truly in contention for the top spot, but it's highly competitive. Then throw in the handful of very serious sub-frontier open weights labs, who are creating huge price competition for the frontier labs, and I have no idea how you can come to this conclusion.
Cloud is no different: 3 major cloud vendors, with 1 clear leader...and now a long tail of tier 2s and neoclouds that are also investing incredible sums to compete.
Where is the pre-structured duopoly? Where is the profit moating?
There are 2 US frontier AI labs. The others aren’t competing at the frontier hence no concern for national security or moating. The moat is 1 main, 1 backup/pusher.
That's brave. I can't help but wonder what dystopian things happen to our data behind the scenes, and what processing/blocks exist to display a different reality of the situation to us.
One scenario could be if they nudge it into answering the above question incorrectly, to encourage the user to correct it, especially where it triggers an emotional response. For example, getting their basic details wrong along with sexuality.
"Rock Solid" Google was known to be privacy invasive. Idk how people trust Google for privacy when it comes to Ai. Look at this thread https://news.ycombinator.com/item?id=40965892
I am in the process of moving my stuff from Google to selfhosted alternatives. Google Photos => Immich, Google Drive => probably Next Cloud or some better alternative.
I am also looking into how to selfhost my email (and yes I know, the horror of inboxing from self hosting. But I am at the point of not only willing to give it a tr, but actively saying "fuck you" to everybody that does not receive my email, because they are on Outlook or O365 (or Google, bt inboxing into Google is easier imho).
Especially with my paid Google now daily having spam in my inbox, while valid emails are so often in spam, that I have to check spam at least once per week.
As someone who self hosts many services, I would encourage you to switch to self hosted email last. Even though self hosting has been made much easier thanks to really incredible FOSS tools, you are going to have random learning pains, some delayed by a few months, and it would really suck to lose email for even the shortest durations. The others are easy to live without while you debug during the learning process.
Btw immich is incredible. I recommend Nextcloud AIO instead of raw Nextcloud. Absolutely great tools
I think the low cost reasonable option right now is to self-host a receiving address and then pick the email provider that looks the least trash and forward your stuff to them to actually handle sending/browsing. It gives you the option to go full self-hosted in the future if your mind changes with minimal headaches but costs practically no effort for the short term.
This is what I do. OpenSMTPD to receive mail, Dovecot to handle IMAP and spamassassin for full spam rules control.
I still configure my clients (Thunderbird, K-9, etc.) to send to my mail server, but that just forwards on to the actual email sender (if authenticated). This way I can change email senders without needing to change all my clients. There are still risks though because if you choose the cheaper plan (as I do), then you get grouped onto a sending pool that spammers might use and then you might get periodically affected if that set of IPs take a reputation hit.
My config for sending: With OpenSMTPD, process it with DKIM:
action "process_dkim" relay host smtp://127.0.0.1:10027
match from auth for any action "process_dkim"
A local DKIM process then processes and sends back to OpenSMTPD:
Then back to OpenSMTPD to forward to the final email sender:
listen on lo port 10028 tag DKIM
action "process_outbound" relay host smtp+tls://label@... auth <secrets>
match tag DKIM for any action "process_outbound"
I'm not sure if I even need to sign with DKIM as my email sender does the final DKIM signing, but it seems that in the email world, it's always best to do things as "properly" as possible whenever you can.
I've been doing this for years, and spam and deliverability issues have been minimal and I've dealt with some big names.
I've spent the last decade building a Dropbox alternative that works on top of any storage protocol (SFTP, S3, SMB, IFS, GIT WebDAV, local disk, and literally everything else): https://github.com/mickael-kerjean/filestash
These days I just don't trust Nextcloud exposed to the public internet. You basically need to be a full time dev ops / security engineer to keep on top of things like PHP2Shell fast enough.
My setup is Nextcloud behind tailscale and if I need to share something I just put it on google drive temporarily.
With a reverse proxy, your only public endpoint is port 443 and nginx or whatever webserver you choose. Sure, the web server might also have zero days but the attack surface is relatively small.
Which was enough exposure to get a shell on your server via the recent php exploit. If you run a public facing nextcloud server you need to be ready to update at least within an hour of patches releasing.
NextCloud AIO runs in Docker containers, so it'd be a jump to break out of that, but, yes, they could exfiltrate your data, though the PHP2Shell PoCs I've seen aren't trivial (unlike the Alpine ssh vulnerability).
I guess it's choose your poison. What are?
P(companies&governments spying on you)
P(hacker stealing your files)
The impacts of the latter are probably more dangerous but unless you're a juicy target, I'd imagine most hackers are going to go for rich grandmas (sadly). So my gut feelings are ~0.95 and ~0.04, respectively.
For me I don't share that many files with others and the ones I do share are usually less confidential than the rest. So I run Nextcloud behind tailscale and put just the share files on google drive.
I might eventually look in to an automatic updates setup for nextcloud to keep up on patches faster than I could do it manually.
But it isn't - not for doing stuff like self-hosting files? It's actually incredibly cheap, most people probably have a commodity PC lying around and can shove a few cheap HDDs into it and RAID stripe it if they're ambitious... or just full mirror their files if they aren't. And if you're really scared of loss find some host provider and rent a cheapo instance somewhere geographically far from you to park a mirror of the data.
I've been a proponent of theirs for years, and I'm still a user. However, their CEO's recent behavior and the public backlash have got me really nervous that there might be a major breach, and that if there is they'll handle it poorly. Failing that I wonder if the company can stay viable with a leader like that.
Vibe coding is one thing, but calling employees losers for caring about quality.. It sounds like he forgot why people use proton.
I had a Google Meet Gemini transcription take well over 24hours once recently, rather than the usual <1hr. I could only assume manual intervention was the cause.
It might also explain why Google do not provide a page to view transcription job queues from meet. You just have to sit and wonder
There are explicit data retention contracts for Workspace (including Gemini transcription) and Google claim this data is never used for training or human review, so this is fairly baseless.
This gist is specifically about the consumer version of Gemini used in Gemini Apps. Basically if you connect your consumer Gemini (which always sends stuff to the training pipeline) to your personal Google Apps account, it gets tools that it can invoke to read your Google Apps data, which then might appear back in your consumer Gemini transcripts that are getting sent to the training pipeline.
Doing this is a stupid product move from Google (per the usual) because it confuses their enterprise customers into thinking Really Bad Stuff like this is happening.
Of course they do. Now that all the freely available training material has been consumed, every website ingested, every image classified, every YouTube video watched, and even books scanned and destroyed, the insatiable hunger for more data must be fed.
Any tool having manual human reviewers or annotator like all of Gemini and notebooklm (now Gemini notebook) suffer from this. Your data has high chance being read by random people or being leaked for training because they might flag it then use it to "improve" the model.
This is exactly why I spent most of this year building the new software stack at https://safebots.ai
I am just fed up with people having no choice but to trust giant corporations.
Yes, I liked blockchain in 2018, because it was the only environment where I could really trust the code. But, it was public, slow and expensive. So I invented something new this year, called the Safebox. Imagine if you could run your entire company from one or more Safeboxes in whatever cloud you want, not have to actually manage anything, but trust auditors of your choice to sign off on the software stack. All your data goes in, and no one can exfiltrate it. Your policies and business rules are guaranteed to apply to every single piece of data. Oh, and it's also nigh-unhackable, unlike your on-prem unpatched Linux and 100 employees' laptops. Wouldn't that be nice?
Note that this is specifically Consumer Gemini with Connected Apps; Workspace Gemini (and Workspace Intelligence, the RAG thing for enterprise/Workspace) still claims not to be subject to human review:
So basically if you use consumer Gemini and specifically authorize the GMail app, for example, then it will sometimes use the GMail tool to load your mail from GMail, and that data (like all data in consumer Gemini) might be given to the human annotation pipeline.
This whole thing seems deeply stupid to me from Google's standpoint, although they've never been good at product management - there's _way_ too much nuance to what's going on and even if they aren't technically affected, this kind of thing will almost surely make their paying enterprise customers leery.
Somebody should start an "exceeds authorized access" computer fraud and abuse case prosecution over this. Perhaps the new Texas state attorney general.
It's really embarrassing that Google does these things, its even more embarassing that their UI / UX is just really awful and all over the place. I've never been in the shoes of having to push a dark pattern into production, but I can't help but feel I'd start looking to work literally anywhere else.
You have to remember that Google is first and foremost a data company. I don't want to sound conspiratorial but every misclick or misunderstood permission window does give Google more data which means more money.
> Google Photos data is used to infer your interests, relationships to people in your photos, and where you've been, including by associating your face with corresponding location data and timestamps. This data is also used to do things for you, like look up your info or add items to your grocery list, and to improve Google services, including training generative AI.
How does that work exactly if the images aren't even real and are generated by AI? When Google siphons up these photos to help them train their AI, the AI is now going to make associations as depicted in that photograph?
That you really were with that person, in that place, in that situation, on that date? As some sort of a fact?
Can someone else create that image and just put it in their library and then have that become a fact on your behalf when picked up, analyzed and trained on?
I have tried to degoogle my life. I already have a way to get my photos to my own server. But how can I take photos on my phone without google getting their fingers into them?
I switched to iOS as the lesser of two evils. Sure, for all their supposed privacy focus Apple does some sketchy stuff too, but at least my personal data is a little bit siloed this way.
I did run CyanogenMod / LineageOS for several years but the reliability I needed just wasn’t there. This may have improved since then, though.
You can use de-googled phones and Immich. I use a sync app ("FolderSync") to upload-then-delete to my server over Tailscale, then Immich picks them up. So I never have storage issues on my phone.
Immich looks like they're trying hard to be a Google Photos clone as far as I can tell. It's a good Android app
I was looking for things like backing up my phone contents to a PC or syncing a music/photo library to my phone and the options I found were brittle python scripts using adb.
FolderSync does exactly what I want and just works. No affiliation just happy customer/user.
Leads to people start calling their senator and start pushing to regulate and break apart every big tech... just as they torrent everything (whoops), they will access everything...
That feeling is your cue to do something about it. Don’t let best be the enemy of good. Automatic backups are great, but if you have irreplaceable data that it would pain you to lose, go get an external hard drive and dump it all on that, right now.
checkout https://github.com/restic/restic you can automate taking incremental backups and shipping them to another location (e.g S3 bucket) which are encrypted by restic so not even AWS can decode them.
You could setup some lifecycles in the bucket as well to prefer deep glacier storage for the cheapest storage tier
As soon as AI started blowing up I rushed all of my data out of Google and to self hosted. It's obvious they were going to suck it all up in to their training and analyze everything to build a perfect profile on you. They always were but the capabilities are supercharged now.
Immich + Nextcloud + Dawarich behind Tailscale has been a dream. I picked up a second hand HP mini PC for cheap, 3D printed a mini server rack to mount it in with some hard drives.
I just use a wireguard VPN with a domain name I already had and DDNS configured, what benefits does Tailscale bring for a situation where only a couple of devices need access?
Most modern AI models carry watermarks embedded in images and text, so they can know if something is AI or not. These are probably detected and discarded from the training set.
You can go grab a model off Hugging Face and install ComfyUI and make as many as you want - right now - on consumer hardware. Not watermarked and very realistic.
And there are techniques to remove watermarks as well.
That is a losing battle on that front. It has to be taken on by not allowing for this behavior that Google is doing. Which I have no solution to.
In the name of the AI race we are taking a lot of shortcuts (privacy, quality, commercial viability etc). As anyone in software knows, the debt will eventually become due.
Google has been doing this kind of shady, privacy destroying, stuff for forever, this has little to do with AI and everything to do what a bad company google is.
So that photo that you sent on email to someone when you thought email was private... it's going to be reviewed and annotated... and trained on it... remind me what stop the thing from becoming the v.001 of the big brother surveillance tool?
That's just irresponsible. If this is anything like old Google search reviews (Appen, etc.) then they will hire anyone as a contractor. I don't want my data being reviewed by some random person.
135 comments
[ 1.4 ms ] story [ 29.6 ms ] threadBut it does explain some conversations with Gemini when I wonder why it's talking about particular things. Like asking it about consultation processes for changes to parks only to get a response that's uncannily tuned to the email I got prompting the enquiry.
No mention of I just read your emails etc .
There’s nowhere to go.
Cloud is no different: 3 major cloud vendors, with 1 clear leader...and now a long tail of tier 2s and neoclouds that are also investing incredible sums to compete.
Where is the pre-structured duopoly? Where is the profit moating?
You can’t build a fighter jet in your garage.
One scenario could be if they nudge it into answering the above question incorrectly, to encourage the user to correct it, especially where it triggers an emotional response. For example, getting their basic details wrong along with sexuality.
Gemini- if it knows anything about you- would be employing some sort of RAG- I didn't mean this as a test of what Google knows about you overall.
Unfortunately, the alternatives are all so bad...
I am also looking into how to selfhost my email (and yes I know, the horror of inboxing from self hosting. But I am at the point of not only willing to give it a tr, but actively saying "fuck you" to everybody that does not receive my email, because they are on Outlook or O365 (or Google, bt inboxing into Google is easier imho).
Especially with my paid Google now daily having spam in my inbox, while valid emails are so often in spam, that I have to check spam at least once per week.
Self hosting email is a nightmare and can bite you in the ass big time.
Self-hosting receiving emails isn't too bad. And you can still control sending too as a relay to the actual sender.
Btw immich is incredible. I recommend Nextcloud AIO instead of raw Nextcloud. Absolutely great tools
I still configure my clients (Thunderbird, K-9, etc.) to send to my mail server, but that just forwards on to the actual email sender (if authenticated). This way I can change email senders without needing to change all my clients. There are still risks though because if you choose the cheaper plan (as I do), then you get grouped onto a sending pool that spammers might use and then you might get periodically affected if that set of IPs take a reputation hit.
My config for sending: With OpenSMTPD, process it with DKIM:
A local DKIM process then processes and sends back to OpenSMTPD: Then back to OpenSMTPD to forward to the final email sender: I'm not sure if I even need to sign with DKIM as my email sender does the final DKIM signing, but it seems that in the email world, it's always best to do things as "properly" as possible whenever you can.I've been doing this for years, and spam and deliverability issues have been minimal and I've dealt with some big names.
My setup is Nextcloud behind tailscale and if I need to share something I just put it on google drive temporarily.
I guess it's choose your poison. What are?
P(companies&governments spying on you)
P(hacker stealing your files)
The impacts of the latter are probably more dangerous but unless you're a juicy target, I'd imagine most hackers are going to go for rich grandmas (sadly). So my gut feelings are ~0.95 and ~0.04, respectively.
I might eventually look in to an automatic updates setup for nextcloud to keep up on patches faster than I could do it manually.
Vibe coding is one thing, but calling employees losers for caring about quality.. It sounds like he forgot why people use proton.
It might also explain why Google do not provide a page to view transcription job queues from meet. You just have to sit and wonder
This gist is specifically about the consumer version of Gemini used in Gemini Apps. Basically if you connect your consumer Gemini (which always sends stuff to the training pipeline) to your personal Google Apps account, it gets tools that it can invoke to read your Google Apps data, which then might appear back in your consumer Gemini transcripts that are getting sent to the training pipeline.
Doing this is a stupid product move from Google (per the usual) because it confuses their enterprise customers into thinking Really Bad Stuff like this is happening.
Clearly the person who bakes my bread should've known better and learned Linux a decade ago to be prepared for this. Sucks to be them I guess.
If only there was some kind of rule about what they can or can't do, which would not be up to them to interpret and enforce.
Something that would make it simpler for regular folks to live their lives without having to rely on trust.
Call it regulation or something.
I am just fed up with people having no choice but to trust giant corporations.
Yes, I liked blockchain in 2018, because it was the only environment where I could really trust the code. But, it was public, slow and expensive. So I invented something new this year, called the Safebox. Imagine if you could run your entire company from one or more Safeboxes in whatever cloud you want, not have to actually manage anything, but trust auditors of your choice to sign off on the software stack. All your data goes in, and no one can exfiltrate it. Your policies and business rules are guaranteed to apply to every single piece of data. Oh, and it's also nigh-unhackable, unlike your on-prem unpatched Linux and 100 employees' laptops. Wouldn't that be nice?
This is my vision: https://safebots.ai/layers.html
https://knowledge.workspace.google.com/admin/generative-ai/g...
So basically if you use consumer Gemini and specifically authorize the GMail app, for example, then it will sometimes use the GMail tool to load your mail from GMail, and that data (like all data in consumer Gemini) might be given to the human annotation pipeline.
This whole thing seems deeply stupid to me from Google's standpoint, although they've never been good at product management - there's _way_ too much nuance to what's going on and even if they aren't technically affected, this kind of thing will almost surely make their paying enterprise customers leery.
It’s not clear what clicking yes does. Is it my entire drive and Gmail? Do I have to give permission per-file/folder/email?
So I click no. And it just gives me the popup again later
It's really embarrassing that Google does these things, its even more embarassing that their UI / UX is just really awful and all over the place. I've never been in the shoes of having to push a dark pattern into production, but I can't help but feel I'd start looking to work literally anywhere else.
I have a strong 'told you so' feeling right now.
> Google Photos data is used to infer your interests, relationships to people in your photos, and where you've been, including by associating your face with corresponding location data and timestamps. This data is also used to do things for you, like look up your info or add items to your grocery list, and to improve Google services, including training generative AI.
How does that work exactly if the images aren't even real and are generated by AI? When Google siphons up these photos to help them train their AI, the AI is now going to make associations as depicted in that photograph?
That you really were with that person, in that place, in that situation, on that date? As some sort of a fact?
Can someone else create that image and just put it in their library and then have that become a fact on your behalf when picked up, analyzed and trained on?
Where does that all lead?
I did run CyanogenMod / LineageOS for several years but the reliability I needed just wasn’t there. This may have improved since then, though.
How many of those photos do you need?
You can use de-googled phones and Immich. I use a sync app ("FolderSync") to upload-then-delete to my server over Tailscale, then Immich picks them up. So I never have storage issues on my phone.
Immich looks like they're trying hard to be a Google Photos clone as far as I can tell. It's a good Android app
I was looking for things like backing up my phone contents to a PC or syncing a music/photo library to my phone and the options I found were brittle python scripts using adb.
FolderSync does exactly what I want and just works. No affiliation just happy customer/user.
But if you have them already on your server then just some tailscale/wireguard perhaps might work for you.
(me neither, but i should)
I wouldn't recommend rsync though, better something that can allow you to get back files you deleted before the last backup.
You could setup some lifecycles in the bucket as well to prefer deep glacier storage for the cheapest storage tier
Immich + Nextcloud + Dawarich behind Tailscale has been a dream. I picked up a second hand HP mini PC for cheap, 3D printed a mini server rack to mount it in with some hard drives.
You can go grab a model off Hugging Face and install ComfyUI and make as many as you want - right now - on consumer hardware. Not watermarked and very realistic.
And there are techniques to remove watermarks as well.
That is a losing battle on that front. It has to be taken on by not allowing for this behavior that Google is doing. Which I have no solution to.
Does HN like censorship that much?