This requires keeping history if, at the time, Buckmaster's account had a certain flag set, because just because the account has the flag now doesn't mean it had the flag at a certain moment in the past. And even if they had such history, it's not obvious whether they just load all data as-is into training.
A totally reasonable pipeline may be unauditable for this purpose.
My employer bought me a Claude Max subscription. On heavy weeks I use 80% of the subscription. And among software engineers that I know, I'm a relatively heavy user.
Why? Because in my experience, the bottleneck is in shareholders approving new features, not my ability to dish out code.
"Age-restriction laws push platforms toward intrusive verification systems that often directly conflict with modern data-privacy law" - when you make rules contradictory, someone always violate these laws, and you can use selective persecution to "convince" companies to favor you, the incumbent politician. You don't even have to use such power, just a "joke" may be enough to send have any rational CEO licking your shoes.
European proponents of "anti-big-tech action" make it pretty explicit - broad discretionary power should be given to executive branch, because otherwise "international corporations" will use "loopholes" (and these "loopholes" are, in practice, explicitly written laws used as intended).
Asynchronous (you don't know when and where you will get Internet/local network connectivity) mesh (there is no gov infrastructure to be trusted) networking in presence of malicious (gov operated) and unreliable (destroyed by gov) nodes is probably among hardest problems known to computer science.
Mining in Arctic is a technical nightmare. Typical alloys become brittle in these temperatures, darkness severely limits operations and/or workplace accidents, transport requires ice-breakers, permafrost is fundamentally unstable (heat from buildings melts it, causing sinking), and you need to pay people a lot to work there.
This is unironically the plan some Norwegian politicians came up with. They were thinking about sending asylum seekers and/or difficult to eject immigrants to Svalbard.
Because almost everyone involved in AI race grew up in "winner takes it all" environments, typical for software, and they try really hard to make it reality. This means your model should do everything to just take 90% of market share, or at least 90% of specific niche.
The problem is, they can't find the moat, despite searching very hard, whatever you bake into your AI, your competitors will be able to replicate in few months. This is why OpenAI is striking deal with Disney, because copyright provides such moat.
Alice changed things such that code monkeys algorithms were not patentable (except in some narrow cases where true runtime novelty can be established.) Since the transformers paper, the potential of self authoring content was obvious to those who can afford to think about things rather than hustle all day.
Apple wants to sell AI in an aluminum box while VCs need to prop up data center agrarianism; they need people to believe their server farms are essential.
Not an Apple fanboy but in this case, am rooting for their "your hardware, your model" aspirations.
Altman, Thiel, the VC model of make the serfs tend their server fields, their control of foundation models, is a gross feeling. It comes with the most religious like sense of fealty to political hierarchy and social structure that only exists as hallucination in the dying generations. The 50+ year old crowd cannot generationally churn fast enough.
OpenAIs opsec must be amazing, I had fully expected some version of ChatGPT to be leaked on torrent sites at some point this year. How do you manage to avoid something that could be exfiltrated on a hard disk from escaping your servers in all cases, forever?
People who have had access to the raw weights of ChatGPT are worth more now (presumably millions in SBC) selling their credentials to other labs. Why bother with trying to leak ? In any case other labs catch-up to the capabilities within a few months.
The model size is probably the thing here. I suspect they took the FAANG remote workstation approach, where VScode runs on a remote machine. After all its not that great having a desktop with 8 monster GPUs under your desk. (x100)
Plus moving all that data about is expensive. Keeping things in the datacenter is means its faster and easier to secure.
Totally agree, people love to talk about how hopelessly behind Apple is in terms of AI progress when they’re in a better position to compete directly against Nvidia on hardware than anyone else.
Apple's always had great potential. They've struggled to execute on it.
But really, so has everyone else. There's two "races" for AI - creating models, and finding a consumer use case for them. Apple just isn't competing in creating models similar to the likes of OpenAI or Google. They also haven't really done much with using AI technology to deliver 'revolutionary' general purpose user-facing features using LLMs, but neither has anyone else beyond chat bots.
I'm not convinced ChatGPT as a consumer product can sustain current valuations, and everyone is still clamouring to find another way to present this tech to consumers.
I think a major part of it is the shovel selling. Nvidia is selling shovels to OpenAI. OpenAI is selling shovels to endless B2B, Consulting, Accounting, software firms buying into it...
> your competitors will be able to replicate in few months.
Will they really be able to replicate the quality while spending significantly less in compute investment? If not then the moat is still how much capital you can acquire for burning on training?
Oh, you mean specifically GPL v3 license, not any GPL license.
Yeah, broad tivoisation and patent clauses make it a problem, because making any patent litigation on unrelated grounds has potential to lose ability to ship the entire OS.
Is that true? If I make a product, and that product runs some embedded Linux system with GPLv3-licensed coreutils, are you confident that my product isn't infected by GPLv3?
Canonical is trying to position Ubuntu as a relevant player in the embedded space.
"Weaponized activists" was an instrument of non-market competition for a long while now.
It's kind of a more modern, more legal take on "send some mobsters to mess them up". You find (or make) an activist group opposing a certain development - and then covertly funnel funding and support to them so that they can do as much damage as possible and stall your competition for as long as possible.
A totally reasonable pipeline may be unauditable for this purpose.