timmg

9 ms·

121.
by timmg 1y ago
The funny thing about bubbles is that they are impossible to predict. My sense is that they always go on longer than is at all rational. (Assuming this is a bubble, who knows.) I’m still waiting for the crypto “bubble” to burst. Seems like
122.
by timmg 1y ago
It would be interesting to know how many queries they get per "base" model. I honestly have no idea what the scale is. In any case, that's the only way to know how to amortize training cost over inference.
123.
by timmg 1y ago
If I'm doing the math right, you can get ~4000 queries per kWh. A quick Google search gave $0.04 per kWh when bought in bulk. So you can get around 100k queries per dollar of electricity. That's.... a lot cheaper than I would ha
124.
by timmg 1y ago
So: as I understand it, Fermet claimed there was an elegant proof. The proof we've found later is very complex. Is the consensus that he never had the proof (he was wrong or was joking) -- or that it's possible we just never found
125.
by timmg 1y ago
> their desire to minimize attack surface trumps any tendency to leave well enough alone. Is that a good thing or a bad thing? Technical people like us have our desires. But the billions of people doing banking on their browsers probabl
126.
by timmg 1y ago
About a year ago, I gave a film script to an LLM and asked for a summary. It was written by a friend and there was no chance it or its summary was in the training data. It did a really good -- surprisingly good -- job. That incident has be
127.
by timmg 1y ago
> Small decrease in employment in exchange for ~25% higher wages for those employed? It's a 25% higher minimum . It doesn't mean everyone was making the minimum before the law. Certainly not all were. (It would be interesti
128.
by timmg 1y ago
96% during a unique event. I think you would typically consider long term in a stat like that. Assuming it was close to 100% the rest of the year, that works out to 99.97% over 12 months.
129.
by timmg 1y ago
It says “usage-based pricing” is coming soon. I think that is the sweet spot for a service like this. I pay $20 to Anthropic, so I don’t think I’d get enough use out of this for the $20 fee. But being able to spin up any of these models and
130.
by timmg 1y ago
Orthogonal, but I just wanted to say how awesome Ollama is. It took 2 seconds to find the model and a minute to download and now I'm using it. Kudos to that team.
131.
by timmg 1y ago
I’ve been curious, for a while, about how our genes affect outcomes. There are kinda two extremes “blank-slate-ism” and “genetic-determinism”. I assume it is always some combination, with a lean in one direction or another. I know the dis
132.
by timmg 1y ago
Does xAI have a similar amount of revenue/customers? (I've personally never heard anyone talk about using it -- except on Twitter.)
133.
by timmg 1y ago
What I really want to know, from someone who does know: Is Intel cooked? Like, will they be able to manufacture chips that compete with TSMC? They used to be a crown-jewel of US tech. But it seems like every time I read the news, the
134.
by timmg 1y ago
Unless they finished the road in the past two years: it's a pain to get to. We went a few years ago and were really surprised it wasn't more famous and had more tourists. I feel like there were about a dozen tourists visiting the
135.
by timmg 1y ago
I'm pretty intrigued by companies like Anduril. I don't know much about them. But the idea that we apply more "tech startup" business models to defense doesn't seem like a bad idea to me. My sense is that the ex
136.
by timmg 1y ago
> The priorities are mutually exclusive: delegating scene data conversion to in-camera engine grants you the most simplicity and ease of use at the expense of control; the territory of technical details grants you the most ability to mak
137.
by timmg 1y ago
Interesting. I think the key to what you wrote is "poorly definined". I find LLMs to be generally intelligent . So I feel like "we are already there" -- by some definition of AGI. At least how I think of it. Maybe
138.
by timmg 1y ago
How cool would it be to directly image artificial light on the "dark side" of a planet (like all the photos you see of lights on earth at night)? I mean, even if there is life it's like 1 in a gazillion. But you could imagin
139.
by timmg 1y ago
> The more Google insists on forcing advertising on us... You can... just not visit youtube, right?
140.
by timmg 1y ago
And now I am too!
141.
by timmg 1y ago
https://vassarrobotics.com/
142.
by timmg 1y ago
Funny, I was just about to build an SO-101, but tariffs adding $100 to the price of the servos annoyed me. How do I buy your kit, please?
143.
by timmg 1y ago
> their revenue comes mostly from Google, they need to diversify Arguably, they get more revenue than what they need to fund browser development. So they could be banking a fair amount. That would allow them to build an endowment that m
144.
by timmg 1y ago
Doesn't give a lot of information about what this is for or how it works :/
145.
by timmg 1y ago
Works now.
146.
by timmg 1y ago
Tangentially related: I recently read The Moon is a Harsh Mistress (by Heinlein). It was a fantastic read (IMHO). And it has extra relevance right now with the AI/LLM progress we are seeing.
147.
by timmg 1y ago
No. It gives two options: Max or pay-as-you-go. I selected Max and it told me I needed to upgrade from Pro.
148.
by timmg 1y ago
Doesn't give me an option to log in with Pro. I updated and restarted. Anyone else have success logging in with Pro? (It says Max or PayAsYouGo).
149.
by timmg 1y ago
My experience is different than yours. I've found it to perform quite well. Completely changes the way I think about coding now.
150.
by timmg 1y ago
(If you know) how does that work? Are the audio/video/images tokenized the same way as text and then fed in as a stream? Or is the training objective different than "predict next token"? If the former, do you think ther

More ›