What about when the open models get to the point that you don’t need these fancy GPUs? It seems inevitable this will happen over time, we see cracks right now within even just GPT w/ expert streaming and flash attention. ASICs are coming soon too.
The article discussed "CYP2D6 inhibitors like Wellbutrin and DXM" and I can offer one data point. An officemate came in with a bad cough one day. I had over-the-counter cough medicine with DXM handy and he took some. He had never taken DXM before; it is apparently not common in his country of origin.
He felt "shaky" and generally terrible after taking the DXM, and almost went to the hospital. I had never encountered anyone having that reaction before.
That particular colleague was also, it turns out, on a pretty hefty dose of Wellbutrin. I never knew there was a possible connection between the two until now.
I think this is exactly the same problem to decide whether any human is conscious, just because they say they are conscious. Humans can act as if they are conscious too, they can even lie to themselves, their whole life, and never discover the truth about their own mind
Cool. I wonder how much additional harnesses would improve the porting process. In my case, I gave more freedom to Fable 5. The model was able to figure things out. I am hearing that the latest models works better with less restrictions. Curious to know the progress on the above.
I certainly would have, had I suggested the problem trivial! I'll keep that in mind for if I suggest things are trivial rather than suggesting I lack understanding.
> You also can't generally "and so on" constrained combinatorial arrangements like this.
I know you can't generally but in the specific case I proposed you can (3 sets of 3, 4 sets of 4, 5 sets of 5, each dice taking one ordinal of each set).
You might want to check that your proposed generalization criticism applies to the generalization at all before suggesting the arrangement doesn't work :)
Urzula turnning of AC for others, but not for her.[1]
Russian helicopters sitting duck when wild fire rages[2]. Afterall these are Spanish property even though Russian made.
They do not want to find solution to this heat wave getting intense year by year other than war mongering. 35K excess deaths this year[3]. Next year it will be far worse.
Keep postponing Ukrain joining EU, while they have suffered imensley.
I would encourage you to explore professions that you might find more fulfilling. I had found myself in a similar situation as you.
Some ideas that I explored were working for a church or non profit, starting an interior design firm with my wife (she’s a designer), advising startups, etc.
2 years later, I’m advising a couple start ups, on the board of two non profits, and back in tech.
I’m not thrilled about being back in tech. But the other ways I’m spending my time help provide me balance.
1. I don’t think that’s a very strong argument. OpenAI and Anthropic don’t buy the vast majority of GPUs they use they rent capacity.
Nvidia could just the same rent those GPUs out for inference and actually have way better margins than they do right now. Antitrust and putting all your eggs in one basket are why they don’t, similar to TSMC.
2. Neither do AI labs. See Anthropic buying TPUs, deploying with AMD. OpenAI on Maia, Cerebras, their own wafers.
Brigadoon is a musical with book and lyrics by Alan Jay Lerner and score by Frederick Loewe.[1] The plot features two American men who stumble upon Brigadoon, a mysterious Scottish village that appears for only one day every 100 years; one man soon falls in love with a young woman from Brigadoon.
- OpenAI claims Astra beats all benchmarks (compared to Fable and Opus, except "Humanity's Last Exam (w/ tools)"): https://openai.com/index/gpt-6-astra/
There has stopped being a formal procedural consequence for OpenAI leaders to declaring AGI, there is a clear (small) business benefit to doing so, and the capabilities of all the frontier models are impressive. So why not declare AGI? It's not like anyone can prove it's not...
Don't be surprised to see other (or even the same) people declaring AGI again and again, as it becomes the best time to do so for different parties.
I didn't read it as dismissive. I think the "Why doesn't this work?" at the end is key, implying they know they're probably wrong, but don't know why. It's pretty common to form and present hypothesis like this, hoping that anyone who knows the actual theory can easily provide a counterexample showing why it's wrong. It's not at all intended to be dismissive.
For a while, I used a firefox extension called "Temporary Containers" which would, as the name implies, create a temporary container for each site you visit. Combined with specifying some custom rules for ones you want to open in a specific container, that was a very convenient setup. But the extension at some point corrupted my firefox profile and has been unmaintained for a while so I didn't try again. I am not sure what happened and it may have been the combination of extensions that was the cause of that. Sadly, the author passed away, if I remember correctly, and it's been difficult to find maintainers for the project so I gave up on that.
You can still install that extension but from the reported issues the functionality and bugs were increasing with every new firefox version. And obviously that setup only helped with the cookie aspects and not necessarily any security issues from javascript and so on.
Maybe I missed it but they didn't delve much into the TAM.
"Half of all flights in the world are under 2 hours." Then they give examples of Fjord-town hopping in Norway, or Island hopping in Hawaii, which seem very niche.
Will they be able to compete in any of the busiest routes? Could it take me Melbourne-Sydney in 1 - 2 hours for < $100? I'd sacrifice some time and $ for an eco-friendly option, but obviously there's a limit.
>I'm confused, is your argument something like "It's too easy so AGI doesn't need to be able to do it"?
Do you possess magnetoreception? a stupid pigeon can "see" the earth's magentic field. why are you blind to it? does a lack of magnetoreception make your intelligence any less "general"
no, you're just blind to it because that's just the way it is.
LLMs are blind to character counting because that's the way they are.
It didn't stop ChatGPT from finding the Jacobian Conjecture counterexample.
Human intelligence and machine intelligence are only going to cross over to a certain degree.
same as plane flight and bird flight are only kinda related.
Ahh yeah, I've seen people using those. I was thinking you'd ideally want to connect them full speed, but it looks like they're 200gbps ports anyway, and I guess even if they were 400gpbs, you don't really need that much for cross-GPU traffic for MoE models doing tensor parallel.
