lumost

Its more of a debate on what the business model is IMO. It's entirely reasonable to be a very large and successful commodity trader/producer, or to make mountains of money on low-margin hive volume transactions.

The current pricing of AI revenue growth is based on them delivering strong software like margins with the growth of a commodity producer. This only really works if you have something sticky which customers will pay a premium for - in the cloud business this became security, automation, "data gravity", and migration costs.

Or you know, NVidia could not charge a 98% margin on their cards...

This isn't really a local model though, it's in the same class as Kimi K3.1 which it is not benchmarked against.

I would find it surprising if the chinese government wanted to allow Anthropic/OpenAI/Google levels of influence on the public opinion and regulatory environment without guardrails. OSS models make AI a replaceable commodity and reduce the potential for market/opinion capture by the frontier labs.

If these all actually worked, I'd presume that there would have been a study on the miraculously healthy 80 year old mice. As it stands I have to assume that most of these mouse studies are fundamentally flawed to begin with and we simply don't check for adverse reactions/overlapping mechanisms/long-term maladies.

TLA+ doesnt prove the code. if they never read the proof, then how do we k ow that it proved the right things?

I think the biggest shift is in predictability. Will something be done today, next week, or next month - nobody knows. Will we converge on a bug free state? not sure.

So far AI works great at fast moving work which you will probably recode or throw away in a month or two anyway (e.g. a startup) but is weak at maintaining a large code base that you rely on. At least part of this is cultural, as the cost of misalignments between team members can be quickly translated into broken software.

What happens if AI is a net loss for the tech industry? It's entirely plausible that AI kills the SaaS business, along with the web and mobile business, alongside large categories of software engineers/tech professionals.

While there might be new stuff, these business's are currently the ones funding AI - if they all get wiped or substantially reduced, then there will be less ai revenue then there is current tech revenue - potentially at reduced margins.

If I understand this clearly enough, what they really did was

A) add a post-processor which accepted a decoded tool call specifying what the processor should invoke.

B) had the post-processor perform the computation in tensor space.

The kv-cache/activations were not shared between qwen/the processor which is how they avoided fine tuning etc. I'm not clear how the tensor representation benefited the computation in this pipeline.

I'd imagine all of the vendors have an incentive to keep their main customer a float, while there might be penalties etc. for missed targets - it doesn't make sense for any of these vendors to ride the contract into chapter 11.

I’m somewhat dubious of the idea that there will be a market for skipped ai operators.

Ai doesn't really give us a new “tool” in as many cases as it “performs tasks.” there isn't any real reason the ai can't perform the supervision task itself. Without a viable market for task doers or supervisors we may observe the equivalent of the 1980s US manufacturing apocalypse in technical white collar fields.

I work in code generation, and often feel the fallacy of “because I can build something that looks 80% like your product in a day, you should pay me N millions.”

It's roughly the same pitch that the ruby on rails crowd had 15 years ago, and the no code crowd 5 years ago.

Very few companies need to rebuild their product every week, or actually care what tech stack it runs on. They are trying to build new things, or reduce how often those things break, or figuring out what if anything is worth building next.

There may be quite minimal reasons for an AI civilization to engage in expansion, with the ability for it to intelligently change and enforce new objectives on itself it could simply choose not to expand beyond resource constraints. Or alternatively invest continuously in greater efficiency improvements.

How much compute could you really pack into a Dyson swarm if that was your only objective? what are the ultimate theoretical bounds for compute density/efficiency? perhaps ai civilizations ride a Moore's law of compute gains indefinitely and simply have no need for expansion.

There are rumors that the employees did not get paid out. Such a deal could be done by, for instance buying the equity from the investors at a current fair market value.

That the new "investor" choses to let their investment die may not matter, it would also be cheaper to selectively buy out investors than buy out all equity in the company.

was it a run around of anti-trust or character's liabilities? It will be interesting if there are creditor lawsuits / trustee clawbacks in the future around these deals.

The suggestion is that this person can get a mortgage to pay their property taxes over the next 20 or 30 years as part of their retirement - or they can sell their property and live well in a cheaper location.

Claiming that the retiree has a right to both the home value and the ability to live there is the friction point. It reads as hollow in the face of rising homelessness and the inability for young people and families to access housing.

I had a thought the other day that you could probably divide up the world into cubic miles and with a few thousand codex subscriptions make a high fidelity blender model of the entire world based on google maps, street view, and tagged location photos.

What would you do with a whole model of the world? I don't know, but it would be fun to walk a national park while I'm on the treadmill or basejump off of a realistic el capitan.

I think this up-skilling sentiment misses that as a worker who thought a seasonal service job on the waterfront was a good deal, up-skilling means taking on 6 figures in debt - after which they will only have a ~1/5 chance of getting a job which compensates for the opportunity cost and debt. For many, this is actually a bad deal and growing their own seasonal business would have been a stronger career path.

The issue is ownership, we have no means of distributing the knowledge from the AI or rewarding those who could help.

We are quickly moving to a world where all symbolic and numeric reasoning for economic purposes is performed by AI.

There is a perceived opportunity cost from someone using a lower-tier model on their task. What if the better model did a "better" job? what if my trials and tribulations are due to model quality?

If you are used to talking to opus5.5 medium, going to GPT6.1 luna low will feel like a step down. Why would any employee take the (personal) risk?

The push for AI adoption in K-8 is the one that strikes me as wrong. The Alpha School and other initiatives are pushing AI on groups for which we have dubious measures of impact. Ostensibly, they are doing this to avoid pushback from traditional curriculum requirements in high school and university - but those kids are going to end up there eventually.

Why would I believe that someone who can't use a mouse/computer is ready to be taught by AI?

My understanding is that the Wuhan Institute of Virology was one of only 2 BSL-4 labs in china in 2019, and the only one focused on emergent viruses. This is out of 684 major Chinese cities.

I recently heard a scientist use a phrase something along the lines of “when two extraordinarily improbably events occur together, there is probably a connection”.

We’ll never know if the lab in Wuhan originated the virus. But it is awfully strange that the first patients of a novel pandemic coronavirus were workers at a lab conducting gain of function research on coronavirus.

I think the big question is whether these recent gains continue. It’s entirely plausible that the ai firms are going to run out of human trainers capable of further refining the model. It’s also possible we are on the cusp of real superintelligence.

Starship’s economics will make such projects inevitable, if it’s not data centers it will be something else. Ideally, the solar panels will be efficient enough to minimize albedo.

A facility in space will likely have better uptime then a terrestrial facility with lower costs for disaster mitigation/insurance. The economics are surprisingly close when you do the back of the napkin kg math, but you can mass-produce orbital data centers in a way that’s not possible for terrestrial facilities.

Granted, the same pitch could have been made for orbital telecommunications vs terrestrial wires… that bet hasn’t paid off yet.

I think the interesting question is whether the average consumer cares? a 4B model should hit roughly the same numbers as GPT5.5 sometime in December/March. That is more than capable of performing a variety of complex work tasks.

The margins on NVidia datacenter hardware are ... high. At least one order of magnitude larger than a consumer chip.

Given the recent deepseekv4.1 advances - how good of a 3B model can we make to run on an iphone natively? is it good enough to match common muse/dot use cases for consumers? the phone is already always on.. no need for a cloud server.

It would be interesting to see how these are justified internally. If I had to guess, it’s either excess budget being channeled into “something” or a justification to reduce street patrols/street crime.

Out of curiosity couldn’t this also mean that we are missing the agriculture? Or that the practices they used were very different from today.

We really need more slow charging stations for folks to just park at. I’d have assumed that the economics would have been favorable to charge by the day/night rather than by the minute.

AI has largely not been trained to operate independently yet. Newer products like grok bot/meta muse point in the direction of an asynchronous "virtual worker" type of persona where the AI is empowered to make at least some decisions.