181–190 of 258 posts
I'm in the other camp: I remember when we thought an AI capable of solving Go was astronomically impossible and yet here we are. This article reads just like the skeptic essays back then. AGI is absolutely possible with current technology - even if it's only capable of running for a single user per-server-farm. ASI on the other hand... https://en.m.wikipedia.org/wiki/Integrated_information_theor...
I'm in the other camp: I remember when we thought an AI capable of solving Go was astronomically impossible and yet here we are. This article reads just like the skeptic essays back then. AGI is absolutely possible with current technology - even if it's only capable of running for a single user per-server-farm. ASI on the other hand... https://en.m.wikipedia.org/wiki/Integrated_information_theor...
What's ASI?
Earlier quoted context omitted.
> It turns out that the larger these models get, the more unexpected emergent capabilities they have, so I'm mostly in the camp of thinking AGI is just a matter of time and resources. AI research has a long history of people saying this. Whenever there is a new fundamental improvement, it looks like you can just keep getting better results by throwing more resources at it. However, eventually we end up reaching a poi…
> LLMs have an additional problem related to training data. We are already throwing all the data we can get our hands on at them. I think it's plausible we'll see a breakthrough in data efficiency that helps here. Humans are an existence proof that language is learnable through much less data and, in theory, facts should only need to be seen once. LLMs in their current form seem very data inefficient.
We probably will see a breakthrough, but there’s no more reason to believe it’ll happen tomorrow than 100 years from now.
-- 1895, Lord Kelvin, president of the Royal Society
> There will never be enough computing power to create AGI using machine learning that can do the same [as the human brain], because we’d run out of natural resources long before we'd even get close I don’t understand how people can so confidently make claims like this. We might underestimate how difficult AGI is, but come on?!
There are less and less people in IT and also Data companies that really care about correctness and efficiency of the solutions. ChatGPT in opinions (not only mine - like I've learned along last few weeks) is getting worse every release. The language gets better, the lies and hallucinations gets better, but being an informative and helpful tool - more like Black Mirrors idea of filling gap after someone who died, not a real improvement in science or social-metrics.
Earlier quoted context omitted.
It is hard for me to not think that given the current approach it will always be almost there, but not quite good enough to be more than a nice assistant or tool that has to be meticulously built and maintained by a company. Take, for example, self driving vehicles, we have been 5 years out for 25 years. That is not to say there has not been significant progress, but to go from a really powerful driver assist to full…
I feel like your point would be better if there weren't already self driving cars around like waymo.
I'm in the other camp: I remember when we thought an AI capable of solving Go was astronomically impossible and yet here we are. This article reads just like the skeptic essays back then. AGI is absolutely possible with current technology - even if it's only capable of running for a single user per-server-farm. ASI on the other hand... https://en.m.wikipedia.org/wiki/Integrated_information_theor...
Earlier quoted context omitted.
I feel like your point would be better if there weren't already self driving cars around like waymo.
That is kind of my point. They exist only in a few select locations. If they were truly self driving Google would be rapidly rolling them out around the World, but they are not.
I think your original post just overestimates the 'normal' pace of change - Taking 10 years to go from the first ride to 100k rides a week is no time at all in the scheme of things.
Airplanes are an example of an insanely fast technological adoption, and they still took 30 years to go from the wright brothers to the first commercial jet. I get the feeling that if the wright brothers invented the plane in 2024 though, people would be saying 10 years after "Planes are all hype! If planes were truly transformative they would be rolled out everywhere and would be carrying passengers by now"
Earlier quoted context omitted.
One thing that keeps me up at night is that the human genome is only 3 giga base pairs, of which only a fraction encodes the design of our brains — and quite inefficiently at that, through layers of indirections. That’s sufficient information to produce a system that can learn to think like us! Not just learn but efficiently , with far less input data needed than any current LLM. Literally just a couple of decades of…
Pretty sure it's not scale but sensors. We have sensors for vision, sound, taste, temperature, texture, etc that are constantly observing the world and affecting not only our current behavior but changing us in real time.
A gazelle just days(!) old can control its body and four legs sufficiently well to outrun a cheetah. This is a complex motor-control loop involving all of its senses. Compare that to Tesla's autopilot training system, which uses many millions of hours of training data and still struggles to move a car... slowly. The equivalent would be a training routine that can take just a handful of days of footage and produce an AI that can win a car race.
There's something magical about neural networks when scaled up to brain sizes. From what I gather, there's little else encoded in the genome except for the high-level pattern of wiring, the equivalent to the PyTorch model configuration.
Earlier quoted context omitted.
That is kind of my point. They exist only in a few select locations. If they were truly self driving Google would be rapidly rolling them out around the World, but they are not.
Waymo's first real driverless ride on a public road was less than 10 years ago. Now they are doing 100k rides a week. I think your original post just overestimates the 'normal' pace of change - Taking 10 years to go from the first ride to 100k rides a week is no time at all in the scheme of things. Airplanes are an example of an insanely fast technological adoption, and they still took 30 years to go from the wright…
45 years: December 1903 to July 1949
https://en.wikipedia.org/wiki/Wright_Flyer
https://www.history.com/this-day-in-history/first-jet-makes-...