> To be clear about where we stand: we believe the world would benefit if the industry adopted a lawful, verifiable, effective mechanism for coordinated pacing as soon as possible.
Can someone explain what coordinated pacing is? I think it's referring to model release but I genuinely have no idea what the authors were trying to say here.
I enjoy Newport myself, he's probably the most level-headed take I've come across. Everyone else has some agenda (or product) they're trying to get at and very much ruins the messaging (ex the agent 'civilization' piece is extremely overblown especially since..that's how multi agent systems coordinate already. Nothing happened that is unprecedented and isn't how the system is designed to work.
I've been doubting him myself (his recent articles just hedge on data centers more than anything else) but then I want to ask, are there any valid critics of AI? Not a "code is bad but it'll get better" but actual criticism in the nature of the financing, politics, etc. I think separating AI as a tool and it's capabilities and AI as product that needs to make a profit is helpful, however most people are really hedging in one camp or the other in their takes.
Assuming this deal goes through, I do wonder what this means for the future of open source models. Like it's very possible the next deepseek model host it's weight on it's site first, vs treating HF as a canonical hosting site. Just my guess, but regardless we'll have to see how labs (and even countries) respond to a less neutral HF.
I'm curious if someone working in DL can speak on what the major things the field is working on for the ~10-15 years? Like I'm curious if another architecture like transformers could be made (in how foundational it was to llms and most ml made today), or what problems should be solved, or may arise, in the future (given a push for ai in science/math I imagine something like verification as an integrated part or logical reasoning).
> The other place I’ve seen tech evangelism is in Haskell, specifically the belief that strongly typed programs are a better way to build software.
Interesting, I'm newer to the field so I say I saw this more particularly with the rewrite in rust crowd, yes rust does have certain aspects better than cpp but does not mean every cpp project would benefit from a rust rewrite. Also is the claim on strongly type programs true? like is it better overall or just better given some specific points?
I know this study is focused in China but one thing I'd like to understand is the nuance in what kind of student is likely to always reach for ai for homework help as I think (especially in the US, can't speak for other countries) the varying powers that be that can determine the quality and type of education you can receive (does your country have means to buy textbooks per student or shared, do you have funds to give students ipads to take home or no, what sets the bar on how far curriculums can go, are teacher and faculty pay and incentives simply tied to student pass rates, etc.) and those do more in shaping what "use AI responsibly" will look like and why it's so different.
> "Could they (Anthropic) get a $2 trillion valuation, yeah they could and I just wonder if it would stay there over time"
Can someone explain exactly how the market can in any sense support not one but two trillion dollar valuations (referencing spacex as the first)? (I imagine openai will likely be in the same ballpark) genuinely we're reaching "elementary levels of big funny number" in the market.
The headline is very misleading, because the manager already knew the employee was tardy, had spoken with privately on it, and merely consulted the AI before the making the decision. Even moreso the AI was faulty on keeping track on attendance. So it's less AI made the right call and AI was second opinion on a clean case already known to all parties. My concern is businesses see the headline and immediately jump to using AI to rank employees, determine who's a good fit for PIP/firing, etc and being able to "well I used the ai responsibly" or "the ai made the same call I would have, it just provided data."
Maybe I missed this, but what are they considering a fact? More specifically, how dated (or recent) should be the answer be? Most of these questions read more like gotchas than trying to understand how and why the model got to it's specifically, especially ones where technically any option was correct depending on the time period looked at something it doesn't seem specified in questions (ie, saying something like "use the most recent data")
Maybe I read it wrong but it seems the case being made is to overhaul the current NSF...with a different NSF? Like I'm failing to see what specifically the current government is doing to hamper AI progress that there should be a "focus on harnessing AI and out-competing China", and what harnessing AI actually means.
I wish the author could have made more concrete examples on this. I guess for me what's more revealing is more recently, the ease to which the leagues (I mainly follow basketball and soccer, but this probably true for many) adopted partnerships with gambling (sorry "prediction market") companies, essentially going from 'we will go after players who gamble to "why don't we let the fans gamble". I think the ease and speed of it more concerning for me.
> In this case, we’re just going to have to hope that this time we make the right choices, for no other reason than that they’re right.
I'm more curious what could be a right choice, and more importantly who is the "we" in this, as many decisions are largely made by companies and governments.
> Garg acknowledges he’s “hard-nosed” and the famous Zoom layoffs severely damaged the company’s reputation — a mistake he knows will continue to haunt him. But as criticized as Garg has been for placing near-impossible demands on the company and its employees, he said Lewis convinced the board he didn’t push hard enough.
> So when shareholders said, ‘You need to take a back seat,’ I complied.”
Man shocked to learn the evolution of shareholder value did not, in fact value him.
Really interesting! I actually use freecad so I'm use to that layer of the process but I'll check this out. Also good to see the work being done in modeling software these days
Can someone in the UK explained how the country has fared with multiple PMs since Borris Johnson? My read as an outsider is it would take another brexit exit (a singular, definable political question that can cleanly classified in a binary label) to create a unifier but I also have no idea what that could be (ai, immigration, climate change, etc are too load bearing and multifaceted to see such a clean binary moment).
Part of the cynic in me just wants to ask "why not make a better harness by default?" The other cynic in me knows I'm about to see a hundred post on 'HOW TO 10X CLAUDE" from the ai bros and I'm already tired.
I guess if I had to ask something (as someone who doesn't use CC as their daily driver), how much control do you have on subagents and roughly how do define or know when a session is getting too long? I know the answer is "when the model is getting worse" but worse is doing a lot of lifting in that sentence.
I guess I'm curious from your perspective what and why makes the tech from being commercially viable, and the concerns of big players adopting this (which was always going to happen I imagine, but when shapes a lot of the future of the field).