Indeed, AKA the OpenAI playbook.

As per usual, in order to understand what it means we need to see :

• performance benchmark (A100 level? H100? B100? GB200 setups?)
• energy consumption (A100 performance level and H100 lower watt? the other way around?)
• networking scalability (how many cards cards can be interconnected for distributed compute? NVLink equivalents?)
• software stack (e.g can it run CUDA and if not what alternatives can be used?)
• yield (how many die are usable, i.e. can it be commercially viable or is it R&D still?)
• price (which regardless of possible subsidies would come from yield)
• volume (how many cards can actually be bought, also dependent on yield)

Still interesting to read after announcements, as per usual, and especially who will actually manufacture them at scale (SMIC? TSMC?).

PS: full disclosure, I still believe self-hosting AI is interesting, cf my notes on it https://fabien.benetou.fr/Content/SelfHostingArtificialIntelligence but that doesn’t mean AGI can be reached, even less that it’d be “soon”. IMHO AI itself as a research field is interesting enough that it doesn’t need grandiose claims, especially not ones leading to learned helplessness.

Read few months ago, warmly recommended. Basically on self selection bias and sharing “impressive” results while ignoring whatever does not work… then claiming it’s just the “beginning”.

I haven’t seriously read the article for now unfortunately (deadline tomorrow) but if there is one thing that I believe is reliable, it’s computational complexity. It’s one thing to be creative, ingenious, find new algorithms and build very efficient processors and datacenters to make things extremely efficient, letting us computer things increasingly complex. It’s another though to “break” free of complexity. It’s just, as far as we currently know, is impossible. What is counter intuitive is that seemingly “simple” behaviors scale terribly, in the sense that one can compute few iterations alone, or with a computer, or with a very powerful set of computers… or with every single existing computers… only to realize that the next iteration of that well understood problem would still NOT be solvable with every computer (even quantum ones) ever made or that could be made based on resources available in say our solar system.

So… yes, it is a “stretch”, maybe even counter intuitive, to go as far as saying it is not and NEVER will be possible to realize AGI, but that’s what their paper claims. It’s a least interesting precisely because it goes against the trend we hear CONSTANTLY pretty much everywhere else.

It’s a classic BigTech marketing trick. They are the only one able to build “it” and it doesn’t matter if we like “it” or not because “it” is coming.

I believed in this BS for longer than I care to admit. I though “Oh yes, that’s progress” so of course it will come, it must come. It’s also very complex so nobody else but such large entities with so much resources can do it.

Then… you start to encounter more and more vaporware. Grandiose announcement and when you try the result you can’t help but be disappointed. You compare what was promised with the result, think it’s cool, kind of, shrug, and move on with your day. It happens again, and again. Sometimes you see something really impressive, you dig and realize it’s a partnership with a startup or a university doing the actual research. The more time passes, the more you realize that all BigTech do it, across technologies. You also realize that your artist friend did something just as cool and as open-source. Their version does not look polished but it works. You find a KickStarter about a product that is genuinely novel (say Oculus DK1) and has no link (initially) with BigTech…

You finally realize, year after year, you have been brain washed to believe only BigTech can do it. It’s false. It’s self serving BS to both prevent you from building and depend on them.

You can build, we can build and we can build better.

Can we build AGI? Maybe. Can they build AGI? They sure want us to believe it but they have lied through their teeth before so until they do deliver, they can NOT.

TL;DR: BigTech is not as powerful as they claim to be and they benefit from the hype, in this AI hype cycle and otherwise. They can’t be trusted.

I’m curious, any advice on that? How does one do “good” telemetry? I’m the first to complain about Microsoft, Apple, (even worst) Google, Meta and now OpenAI collecting data to sell me stuff… but it’s true that also some data is needed to get some kind of introspection in terms of usage. Developers need to understand what is actually happening with the software they develop.

Now I’m wondering specifically about 2 side :

• how to do the data collection correctly (e.g local only, only send on crash, only send without PII, store only aggregate)
• how to get informed consent from users (e.g off by default, UX that supports understanding of why it’s done and how)

I’m genuinely glad that the mindset around privacy have changed since the last few years but I’m wondering how, when it’s a genuinely positive good case (to truly make better products), to do it.

I forgot the exact number but while installing Debian (Bookworm and Sid) this weekend I was shocked by how small the base install, with a window manager (“big” one by your standards, i.e KDE), was. Maybe 2Gb, definitely less than 4Gb. It all worked fine, I could browse the Web, print, edit rich text, watch video, etc.

I installed a ton more stuff since, e.g Steam, Inkscape, Python libraries for computer vision, etc and it’s still not even 10Gb.

So… my suggestion is the same as I shared earlier in https://lemmy.ml/post/20673461/13899831 namely do NOT install preemptively! Assuming you have a fast and stable connection I would argue stick to the bare minimum and all add as you need.

In fact… if you want to be minimalist I would suggest to do another fresh install (it’s fast, less than 1hr and you can do something else at the same time) and stick to the bare minimum right away.

TL;DR: don’t get rid of, just avoid adding from the first place.

It’s a tricky situation to navigate.

There is the technical aspect, namely is it actually feasible, but itself wrapped within an economical and political context, as I’ve highlighted in another thread on this post.

On one hand we learn from Snowden’s leaks about an entire surveillance apparatus, we might also have a conceptual understand of limitations via articles like “On trusting trust”, plain incompetence and shortcuts for large companies, so all that and more invite us to be very prudent. Those are actual justifications for questioning what hardware, if any, can be trusted.

Yet… one can’t go from those justifications to speculate. Yes there might be flaws, intentional or not, in both the design or the production or both of chips. Still, it’s not because it’s conceptually possible, or even that it happened before, that it does happen today and at scale.

Your System76 is an interesting example and it’s a bit like my Banana Pi tinkering, or even more limited (yet exciting IMHO) the Precursor. Namely it’s a very costly trade off today to “work” with hardware one can (at least try to) understand better, hopefully itself leading to better privacy and security. In the end most of us believe the trade off for more affordable performances trumps that deeper understanding.

I must express myself quite poorly. It is not a point about technical knowledge, in fact if you were to know more about the topic than I do, I would expect you to even more be upheld to higher standards and thus not promote a bad solution, even more so assume it’s the only one. I can’t imagine that even a PhD student who is supposedly at the frontier of knowledge in their very narrow field would assume no alternative is possible, or will ever be. This even more the case without having both a complete understand of the landscape but also about OP’s actual needs, which is probably hard to express clearly and thus leading to a lot of assumption. Here maybe a simple loud alarm from a BT speaker going out of range might be enough.

My whole point is that abandoning hope, and leading others to do so, is worst than actively finding for a barely OK compromise.

Anyway I don’t want to invest more energy on this discussion unfortunately so simply wishing you the best, thanks for the clarifications.

I imagine it’s like everything else, you can only realistically verify against a random sample. It’s like trucks passing a border, they should ALL be checked but in practice only few gets checked and punished with the hope that punishment will deter others.

Here if 1 chip is checked for 1 million produced and there is a single problem with it, being a backdoor or “just” a security flaw that is NOT present due to the original design, then the trust in the company producing them is shattered. Nobody who can afford alternatives will want to work with them.

I imagine in a lot of situations the economical risk is not worth it. Even if say a state actor does commission a backdoor to be added and thus tell the producing company they’ll cover their losses, as soon as the news is out nobody will even use the chips so even for a state actor it doesn’t work.

They asked for an alternative to airtags. I provided one.

And even though I’m not OP I’m genuinely grateful for that.

Doesn’t matter if they were compromised because like I said, everyone is eventually.

No! That’s the whole point of this Privacy community! If someone is using, using home automation as an example, Apple HomeKit or Roomba or Google Home they will eventually get compromised BUT if they are using something local, e.g Zigbee with HomeAssistant they WILL never get compromised because by the very local only architecture of that solution no data is leaving the home and thus can NOT be compromised.

The ENTIRE reason d’etre of this community is not to say “Oh well… the default solutions are imperfect, we have to shrug and accept the statu quo” but rather provide genuinely alternative.

I understand a lot of people can enter into a learned helplessness mindset imagining that only poor solutions exist and thus, better pick the least worst one, but by doing that we are giving power to Big Tech, surveillance capitalism, etc.

Please do NOT say that “everybody gets compromised” when you actually mean that “the vast majority of people who accept to use a popular solution with trade offs that are not good for privacy”. It sounds like a finicky difference but it’s actually totally different because it shows that it’s not inevitable.

By taking shortcut in your language you limit what’s conceived as possible by others who are asking for help, again, in a Privacy focused community.

True yet still not OK.

That’s also why a lot of us do try to avoid, as much as is realistically feasible, to provide any data to any company that should store it. Hence why a lot of questions here are about self hosting, no cloud, etc. It’s not paranoia, it’s because companies cut corners and as you correctly point out, fail to keep us safe. So it’s not about Tile specifically, they are just yet another poor example. Let’s not defend them nor this kind of practices. If people in the Privacy community are OK with that, we have a rather deep problem.

The same way you would do it with a black box while optionally taking as many shortcuts as one is comfortable with by virtue of assuming having a better understanding of it’s been built?

Get it audited by tools, e.g OneSpin, or people, e.g Bunnie, that one trusts?

I’m not saying it’s intrinsically safer than other architectures but it is at least more inspectable and, for people who do value trust for whatever, can be again federated.

I assume if you do ask the question you are skeptical about it so curious to know what you believe is a better alternative and why.

Buying other hardware that you (well… not me ;) can inspect and verify, e.g RISC?

For now the performances are pretty terrible BUT one can imagine, assuming they have the right discipline and mental model doing what’s actually personal on a verifiable processor, e.g browsing and reading emails, and what’s not, e.g watching a TV show on another machine with CPU/GPU with an unverifiable architecture.

PS: I have a Precursor and a Banana Pi BPI-F3 with SpacemiT K1 8 core RISC-V chip and that’s the main idea behind them both, i.e knowing, as a community, how it works all the way down.

I… agree but isn’t then contradicting your previous point that innovation will come from large companies if they only try to secure monopolies rather than genuinely innovate? I don’t understand from that perspective who is left to innovate if it’s neither research (focusing on publishing, even though having the actual novel insight and verifying that it does work), not the large companies… and startups don’t get the funding either. Sorry if you mentioned it but I’m now confused as what is left.

They just provide the data. They can question the methodology or even provide another report with a different methodology but if the data is correct (namely no fabricated) then it’s not up to them to see how it’s being used. The user can decide how they define startup, i.e which minimum size, funding types, funding rounds, etc. Sharing their opinion on the startup landscape is unprofessional IMHO. They are of course free to do so but to me it doesn’t question the validity of the original report.

Neat.

Warning disclaimer : I’m not a cryptographer.

I actually tinkered with https://github.com/open-quantum-safe and it’s actually quite simple to become “post-quantum” whatever. The main idea being that one “just” have to switch their cryptographic algorithm, what one uses to encrypt/decrypt a message, from whatever they are using to a quantum-resistant (validated by NIST or whomever you trust to evaluate them) and… voila! The only test I did was setting up Apache httpd and querying that server with Chromium and curl, all with oqs, while disabling cryptographic algorithms that were not post-quantum and I was able (I think ;) to be “safe” relative to this kind of attacks.

Obviously this is assuming a lot, e.g that there are not other flaw in the design of the application, but my point being that becoming quantum-resistant is conceptually at least quite simple.

Anyway, I find it great to demystify this kind of progress and to realize how our stack can indeed, if we do believe it’s worth it now, become resistant to more threats.

Research happens through university, absolutely, and selling products at scale through large companies, but that’s not innovation. Innovation is bringing new products, that is often the result of research yes, to market. Large companies tends to be innovative by buying startups. If there are no startups coming from research coming from universities to buy, I don’t see how large companies, often stuck in the “innovator dilemma”, will be able to innovate.

Thanks for linking to criticism but can you highlight which numbers are off? I can see things about ByteDance, Ant group, Shein but that’s irrelevant as it’s not about the number of past success, solely about the number of new funded startups. Same as the CEO of ITJUZI sharing his opinion, that’s not a number.

Edit: looks totally off, e.g “restaurants, in a single location, such as one city, you could immediately tell that there were large numbers of new companies.” as the article is about funding, not a loan from the bank at the corner of the street.