I tend to agree, but the Semianalysis + Dwarkesh side of reporting still does surface interesting information. You just have to take it all with a grain of salt, as indeed it is more hype focused, and look for real information hidden in the noise. And possibly to be a bit more entertained as you do so.
I suppose access journalism does have to get something out of the bargain, however small it's not zero. If that's the way you like to spend idle time, well it takes all kinds. I don't understand competitive scrabble either.
The trouble is the article is so poorly written I don't want to look for hidden information, i want to close out. I'm guessing ai wrote it, the information is constantly repeating, and not even always consistently
It seems they had a head start but are now facing stiff competition on all fronts. Software moat, GPU's for gaming, and its distant cousin datacenter compute. They rightfully invested their insane profits into many ventures, and how many of those have turned around into profit?
They are also a robotics AI company with Omniverse. They are also an AI company with Nemotron. They are also a bleeding edge network equipment company after the Mellanox aquisition.
They stand to make a lot of money if they succeed in every venture. Good for Jensen taking risks and driving innovation, I hope they succeed in chewing even 50% of what they bit off.
> They rightfully invested their insane profits into many ventures
This is the inherent downside of such a rapid rise to being the world's most valuable company AND still being considered a growth stock. At their massive scale, the number of new adjacent businesses that have both sufficient size and potential growth is limited.
> I hope they succeed in chewing even 50% of what they bit off.
Anything approaching that is vanishingly unlikely. They're being forced to play the game more like a VC. The question is if a few unicorn winners can offset dozens of losers. The challenge is that, unlike a VC, their bets are much more correlated around AI.
That is a truly terrifying future to imagine. Fortunately I would hazard to guess the medical technology to accomplish something like this is very, very far away. Not only because of the raw technological challenge, but also the barriers to development that scientists, doctors, and engineers would face before even being allowed to conduct experiments.
I do agree though the humanoid form is a dead end for robots. Just build giant cubes that process inputs and give outputs, like a dishwasher. Why wash dishes with meat wand tentacles or try to recreate meat wand tentacles when you can accomplish the job in a wholly different way with far greater efficiency...?
Where's the clothes foldeing cube? Analogous to the clothes washer and clothes dryer.... the clothes folder...
Why stop at dishwashing...? Sell an entire integrated robotic kitchen.
Dedicated machines can obviously solve an issue more efficiently, but it’s still good to have a machine that can do most things adequately.
As an example, you can have an automated washer, dryer, and folder, sure. But what if you wanted to automate the retrieval of dirty laundry and the delivery of clean laundry ? That would need to be some sort of robot to travel throughout an environment (fit through human sized areas, open doors, walk steps) to collect and deliver things. And if I have a robot roaming around the house, I would prefer to just buy one robot that could do many things rather than have to buy it to just collect things and more expensive machines as well.
I think the only real barriers are ethical. I'm pretty sure we'll figure out to do a completely synthetic biological robots without any kind of biological brain in the next 10-20 years, and they don't even have to resemble animals or humans. This should alleviate some of the ethical issues.
The author of the article mainly talks about agentic programming, code generation, and reasoning. Ans very rightly identifies a big problem with agentic programming, in my experience. If developers can't maintain the software without AI, it's doubtful they can steer AI to maintain it either. Maybe this is not true and we can tell AI something like 'reduce the number of lines of code' until the essential software is exposed and pared down to a quantity and modularity that humans can then participate.
No doubt, most of the value creation is outside of creating software. But if Nvidia and Anthropic do succeed in making better hardware and better software, then the positive reinforcement loop does seem like it could take off. And coding is a big part of that.
Maybe we don't need to understand the code at all? Hard to fathom.
> Maybe we don't need to understand the code at all? Hard to fathom.
Doesn't seem too crazy. This is the life of every customer (and most managers) of a software dev. To fathom it, think of yourself as the unicorn customer that can define projects, specs, and methods with competence, and even make architectural calls, with the AI being the dev.
To me, this is the incredibly clear eventuality of our profession, within the next 5-10 years. You writing code will be the same as most devs who try to hand write ASM: you will very very rarely be able to beat the compiler at its own game. And, just as now, there's a very small chance you'll go read what compiler produces.
I'm a decent full stack web developer, who knows almost nothing about kernel architecture or machine code. Before LLMs I could make a functional app, despite not understanding several layers the app ran on top of.
Determinism is NOT a requirement, as proven by your existence of software devs. Again, shift the perspective to the people you work for, or to your fellow dev. They get along fine with you, who are also not deterministic. They check your output by setting requirements that you can prove are true in a way that satisfies them. You yourself guarantee correctness by tests, because you learned long ago that you yourself are not to be trusted, and are not capable of writing 100% correct code on first pass, except for the most trivial tasks. Your lack of determinism is built into your workflow, with tools like linters, code analysis, etc, to help deal with it all.
Do you think this necessarily sends us as a society (as a species) towards a future where we eventually stop understanding technology and science itself?
Like in Asimov's Foundation or Liu Cixin's 'Taking Care of God'.
Maybe we need two tiers of technology. Human in the loop up to the point of AI, and then the AI layer which depends on the human stack.
As knowledge progresses, less and less people will be able to understand it, as has always been the case with the progression of technology. This is an inevitability since intelligence is something that can only be slightly learned [1]. The passionate, high intelligence, nerds will continue to do what they've always done, being the only, very very small, percentage of society that bothers to understand any of it.
Most humans are dumb, from a science perspective. The ability to be successful in science has an intelligence threshold of somewhere around 120 [2], with true innovation requiring much more (as history has shown).
Do you see any alternative to 'Taking Care of God', if we continue progressing?
But, for coding specifically, coding as an unfortunate consequence of working with computers. We've been increasing abstraction since their inception, so we can do the actual work with them that we're trying to implement with code as a means to implement it. Go ask and older programmer, and he'll say the same sort of "kids these days don't understand computers", because he had to wire up the logic on a breadboard and write machine code in a hex editor, where you use a javascript library without even knowing how to implement an adder!!!
I take your point, statistically we're already there.
Unless humanity was to bootstrap its own biological intelligence to be consistently higher, there's little point in any individual attempting to safeguard the tech stack to first principles. Accept AI or don't accept AI, history tells us that if it works, people will move to the higher abstraction layer.
I can imagine an organization that devotes itself to preservation of the history of technology... like monastic groups that copied texts endlessly. In a post-scarcity world, that would be a sensible outcome... people are free to devote themselves to art or science or debauchery as they please...
Not really. It is true that supply chain attack concerns has put increased pressure on compiler makers to make compilers deterministic, but historically they haven't been, generally speaking. It was long thought that determinism didn't matter. Where compilers do differ from LLMs is that compiler output correctness is binary, while LLM output correctness is probabilistic.
However, compilers have long suffered from inputs not being correct, so we have put a lot of effort into finding ways to validate that the input is correct relative to the output. Which turns out works just as well for LLMs, to the point that validating that the input is correct relative to the output ends up offering the same assurances for both compilers and LLMs alike. Compiler output correctness alone is not actually all that beneficial in practice because, again, it is only as useful as the correctness of the input and inputs are not trustworthy without "double entry accounting" to go along with it. Once you have "double entry accounting" in place then you already have what you need to turn LLM correctness into being binary as well.
I have great interest in fine-tuning open models, and I'm looking for resources that HN folks can personally recommend. This article looks good and I've bookmarked it to read more thoroughly over time.
I've gotten as far as running Nemotron-3-Nano 30b locally, and plan to target models around 30b - 120b parameters. Based on brief examination of the results I can get, I think these vanilla models are capable enough to add real value, but training could push them over the finish line for specialized tasks.
What I really appreciate is that the author is thinking about the whole process, which is also my goal. Confirmation that others are identifying the same use case, and the same strategy for adding value using this technology.
This is a long term project, so I plan to buy hardware to conduct the fine-tune. ..
What kind of hardware are you using and what is your token generation speed (tok/s)?
Every time I've tried to run local models, even on my beefy mac book (128Gb Ram), I've been very disappointed at tok/s speed.
I've been using dual NVIDIA GPU's with Ollama. Even when I push to 30b models and more context window, Ollama manages RAM / VRAM very well. I never got OOM errors, just massive slowdown as the PCI-E bottleneck throttles the GPU's. I've had some large prompts take 30 minutes to process.
But I should mention: I am trying to implement workflows and processes akin to CI/CD that run 24/7 in the background. These are not interactive use cases, so I don't care so much about tokens per second.
Sure, most of the developed world is facing a demographic crisis. But will it trigger a meaningful reset of wealth distribution? Or be ineffective like the peasants revolt...
The core question seems to be whether the economic systems and fiat currencies can survive the next century. Population decline and a shrinking economy should go hand in hand... but planning contractions does not seem to be priority...
I maintain some FPGA code in VHDL and also some C code for microcontroller. One thing I appreciate immensely about VHDL is how strict it is. You have no option but to think carefully about modularity and heirarchy. By comparison C is very free and simply offers the opportunity to make your own constraints.
Notably the mental model of the hardware is different for each case. Other species of languages like Lisps and Golang and SIMD fascinate me because computer hardware fascinates me.
I like the toy 2D textual experiment the author has given us, and I agree with you it would not scale. Since the author calls their toy 'goofy' they probably agree. But I would contest (edit: your point that we are not stuck in our ways.) We are by and large seriously stuck in our ways. Even if a perfect visual/spatial programming language dropped into our laps today, I posit it would take over 50 years to meaningfully supplant C et al..
I understand Gates has also helped in reviving Nuclear power, from reading news on this site and others. Smaller, updated designs that don't face quite the same level of pressure from regulators.
If we assume you are right about billionaire philanthropy being basically ineffectual (I personally agree) there is a line of reasoning that I find explains why adequately. When systems don't have their incentives structured properly, then quite often the unexpected outcomes are stronger than the predicted outcome. Because the input to the system did not properly account for, or change the incentives which drive the dynamics of the system.
Examples about in healthcare, social programs, education... large SWE companies...
There's so little real pressure for results when you're backed by some billionaire's fortune, the existence of the organization is not threatened by non-performance... there's no free market to survive in, the goal is to lose money... the things you are trying to measure are slow signals or mostly qualitative...
