Closi

37 ms·

601.
by Closi 3y ago
> AI will never be better at avoiding a playing child who accidentally stepped on the road as an average human driver. Very hard to predict what will be possible with AI in 10 years time, but I wouldn’t bet on humans here. Humans have a
602.
by Closi 3y ago
Apologies I did misread - removed original comment.
603.
by Closi 3y ago
Edit: Misread and assumed 1100 euros / 6 months. Have removed comment
604.
by Closi 3y ago
I can’t believe my M1 Mac is almost 3 years old, performs just like the day I bought it (blazingly fast). My work Thinkpad from the same period feels half way dead.
605.
by Closi 3y ago
> the only valid values for collection sizes are 0 and N -> ∞. Of course there are valid values below that... If they were designing to a collection size of N -> ∞ then they shouldn't have used Insertion Sort either then, bec
606.
by Closi 3y ago
I don't think the 90's developers could have forseen that 2ms on boot would be huge for cloud serverless compute like lamdba.
607.
by Closi 3y ago
Yeah, but it was working. And maybe it wasn't 'no additional effort' because they were more confident with quickly writing a bubble sort.
608.
by Closi 3y ago
I think we have different definitions of stupid - the original implementation was perfectly fine for the time and requirements it was written for.
609.
by Closi 3y ago
There is a different pressure on refactoring in 2023 than in the 1980s / early 1990s. If you need to fit your application into 5kb of pascal code so it can be bundled with an OS, there’s going to be less code to refactor and probably l
610.
by Closi 3y ago
Anecdotally my experience dealing with German companies being in the UK is that they can be highly detail-oriented and process driven, to the point where they can be hard to work with if you are outside Germany. It seems that progressing fa
611.
by Closi 3y ago
What do you think LLMs are made of?
612.
by Closi 3y ago
They aren't saying slavery is good, they are saying it's an extreme example of something that isn't actually an objective fact. More subtle examples would be gun-control or abortion, where two intelligent people can hold wh
613.
by Closi 3y ago
And computers are just lots of really fast logic gates. I think the issue with reducing LLMs to "next word predictors" is that it focuses on one part of the mechanics while missing what actually ends up happening (it building lots
614.
by Closi 3y ago
> If we consider that GPT-4 is about as capable as a reasonably intelligent high school graduate or college freshmen, I wonder how much actual benefit we get from multi-agent setups. It's worth noting that GPT-4 internally uses a Mi
615.
by Closi 3y ago
Probably about as good as I would do if I was hand-coding a SVG and was given an hour. People forget the thing we learnt during the computing revolution - computers don’t need to do individual calculations that humans cannot, they just need
616.
by Closi 3y ago
That’s true, but if your argument can be extended to absurdity it probably can't be used to justify where the 'correct' place on the scale is to land.
617.
by Closi 3y ago
The tahitans can deal with communication just fine with only 13 letters, so I guess we could probably trim the letters b, c, d and g out of unicode too. With enough optimisation we might just be able to communicate with just a boolean-based
618.
by Closi 3y ago
On a global scale it's tiny, but on a regional scale it could really impact profitability. Assuming 85% of Norwegians have an account: * 98,500 a day = $35 million per year * $35 million / (5 million Norway population * 85% assump
619.
by Closi 3y ago
It sounds to me like the novel view is generated from another algorithm - then their algorithm makes a mesh from these generated novel views to 'improve' the output in a two stage process (coarse -> fine).
620.
by Closi 3y ago
They are using the word to mean that it is not single-purpose or task-specific and there is an ability to fine-tune / adapt it for new use cases without retraining the whole thing. > With additional fine tuning, the base model can b
621.
by Closi 3y ago
I agree with your maths - It’s possible they need more error correction bits or something like that, and there is a reference in the article to needing more bits to account for “chromatic abberations”
622.
by Closi 3y ago
Or better yet - 'Open Scene Description'. For the extra two syllables they can say what it is.
623.
by Closi 3y ago
By easier they probably mean “it’s easier to do the same thing” - eg centering a div. It’s also become deeper at the same time though - so it might be easier to do the same thing and harder to know it all.
624.
by Closi 3y ago
I've got the username 'admin' on imgur. I constantly get password reset requests from people trying to get admin access to imgur.
625.
by Closi 3y ago
Personally ChatGPT generally gives me a quicker, better, simpler and ad-free result for the snippet (At least with GPT4). Plus I can ask follow-up questions in a context-driven way ("Can I do this without importing a library?"). I
626.
by Closi 3y ago
> I haven't. Because (free, as in free beer) chatGPT is extremely slow, I have to make a rather extensive proompt to get the result I want to, and then I still have to debug most code. That's because you are comparing asking Ch
627.
by Closi 3y ago
Much like half the answers on stack overflow. Rather than ask it to write a full ansible job, try asking it the question you would ask on Stack Overflow (i.e. you probably wouldn't ask Stack Overflow to write the full job for you).
628.
by Closi 3y ago
It now redirects to twitter for me, so I guess it was something in the chain that hadn’t refreshed!
629.
by Closi 3y ago
They made the change but https://x.com still leads to a GoDaddy parking page for me…
630.
by Closi 3y ago
I don’t think this is quite right with all conferences - I’ve personally found it quite common for these event companies to give talks to “platinum” sponsors as an indirect way of charging for them. (Although this particular instance sounds

More ›