summerlight

no profile record

comments

11 gün öncediscuss

The title is kinda misleading. The actual wording is not specific about humanoid robots but "physical AI" which encompasses every machinery that can be potentially integrated with AI, especially focused on mass manufacturing for the Korean case. Basically this project is about all physical infrastructures to automate high tech manufacturing industry.

11 gün öncediscuss

Looks like a lazy translation; the president used a word "대도약" while the Chinese campaign that you're referring is translated into "대약진운동".

20 gün öncediscuss

My personal rule of thumb: I am usually okay with agents driving e2e implementations if this won't make life noticeably worse when it does not work. Some analytical code? Perfectly fine. Hobby projects? Fine, though I prefer doing a fun part myself. Refactoring production code generating 10x more revenue than my salary? You'd better be at least understanding what it does.

2 ay öncediscuss

> 99% of the server code doesn't deserve trade secret protection,

Unfortunately, it is not you define whether it deserves. It's a court, ultimately SCOTUS.

2 ay öncediscuss

This is one of the reason why that law will surely be challenged and very likely invalidated by SCOTUS. Trade secret protection is a very fundamental part and if this is forced to be broken by legally compelled speech, then it needs to have very creative interpretations over judicial precedents.

2 ay öncediscuss

Many developers with good will actually tried that and gave up due to lots of problems. This is not just bad ROI but also a legal minefield. Engineers usually cannot argue against this kind of risks. Enforcing this will unlikely work in higher courts. Though something like open sourcing protocols for server reimplementation may have some chances.

While I see problems in the law but the spirit is reasonable. We need to push toward this direction. At least there should be difficult economical trade-off for publishers when they decide to shut down the game. Nowadays, some random executive just takes look into some excel, see some games have declining revenue and decides to "simplify the business" without much thoughts. This has to stop.

2 ay öncediscuss

Because of high coupling dependencies between google3 projects, compiling just a single project usually pulls hundreds of thousands of different build targets.

2 ay öncediscuss

All the developments are done in a virtual remote file system. From editing to compiling, everything is done remotely. Of course this does not fully stop people from doing manual c&p, but it still makes it hard enough to discourage it.

4 ay öncediscuss

I guess it is much more frequent to maintain internal patches rather than doing all the merging work into upstream, especially the feature is non-trivial. Merging upstream consumes more time externally and internally, and many developers are working with an aggressive timeline. I don't think it is fair to criticize them because they didn't do ideal things from the beginning.

5 ay öncediscuss

The quality of generated code does not matter. The problem is when it breaks 2 AM and you're burning thousands of dollars every minutes. You don't own the code that you don't understand, but unfortunately that does not mean you don't own the responsibility as well. Good luck on writing the postmortem, your boss will have lots of question for you.

6 ay öncediscuss

This is not very surprising. I've always thought that it's more of correlation than causation. If you're a good problem solver, then there is a good chance that you are probably good at both college admission and software engineering. So companies have been using it as their proxy for hiring because... why not. I'm not saying college curricula are useless, but this dependency on (imperfect) correlation might have caused significant opportunity costs for talent acquisition and now companies are slowly acknowledging it.

6 ay öncediscuss

Yeah, meet and chat (where each has their own bloat) are now integrated into the mail app as well. This contributes a lot.

7 ay öncediscuss

I guess if they want to eventually deprecate the 2.5 family they will need to provide a substitute. And there are huge demands for cheap models.

7 ay öncediscuss

I just hope them to provide an option to get rid of all those predictive models and just use a static, consistent layout. At least I can blame myself if my typo is from my own mistake.

7 ay öncediscuss

This looks interesting. This project has some novelty as a research and actually delivered a promising PoC but as a product it implies that its training was severely constrained by computing resources, which correlates well with the report that their CFO overruled CEO's decision on ML infra investment.

JG's recent departure and follow up massive reorg to get rid of AI, rumors on Tim's upcoming step down in early 2026... All of these signals indicate that those non-ML folks have won corporate politics to reduce the in-house AI efforts.

I suppose this was a part of serious efforts to deliver in-house models but the directional changes on AI strategy made them to give up. What a shame... At least the approach itself seem interesting and hope others to take a look and use it for building something useful.

8 ay öncediscuss

Don't forget to mention automatic enrollment of your production group into access-on-demand. Any minor access on the production now requires the group manager's approval. I had a fun time with some production fire where only director level folks can approve the access. Even funnier thing is that this "refactor" was done without any prior notice.

9 ay öncediscuss

Basically it all boils down to budget. Those engineers knew this is a problem and wanted to fix that but that costs some money. And you know, bean counters in the treasury are basically like, "well it works well, why do we need that fix?" and the last conservative govt. was in a full spending cut mode. You know what happened there.

9 ay öncediscuss

Internally, TPU is much cheaper for the same amount of compute compared to GPU, so I don't see much reasons why they need to use GPU. Probably >99% of compute budgets are spent on TPU. It might be true if you say these <1% still counts, but I guess it is pretty safe to say all of its meaningful production workload are running on TPU. It is simply too expensive to run a meaningful amount of compute on non-TPU.

Just to clarify, TPU has been in development for a decade and it is quite mature these days. Years ago internal consumers had to accept the CPU/GPU and TPU duality but I think this case is getting rarer. I guess this is even more true for DeepMind since itself owns a ML infra team. They likely be able to fix most of the issues with a high priority.

10 ay öncediscuss

> So do a currency swap?

It is the US who refuses to do.

10 ay öncediscuss

One condition that Washington has been broadly mandating is, when they ask for the money then it should be deposited in the treasury's account within a short period. Basically they control the pace of the investment and the detail is not even a part of the deal.