Kaggle has a 10 million dollar competition running now,
Currently the AIs are solving 18 of 50 trial problems. There is a 10,000$ prize for the first shared notebook that solves 20 of the problems.
It doesn't look like it emitted 'training data' - but rather random strings that matches personally identifiable information.
If you have ever tried to generate a new email address you will frequently encounter your chosen username already existing. Same thing - a LLM trained to generate 'email like' strings, will naturally generate random strings that collide with real addresses even if the address didn't exist in the training data.
trading compute for equity is not 'free'
Almost all of the inflation is due to a handful of causes -
1) oil producers cutting supplies and US gasoline refineries closing plants - this has caused a continuous spike in gasoline prices. Gasoline has a poor elasticity of demand (people have to go to work, school, etc. and have no choice but to buy; businesses have a fairly fixed gasoline requirement for things like deliveries, etc.)
This has a knock on effect on all prices of anything that requires transport.
Similarly anything requiring natural gas or oil in the manufacturing process is more expensive (any heating, any plastics, etc.)
2) Car manufacturers cutting orders of microchips causing a shortage of new vehicles causing new and used car prices to surge. Again there isn't a ton of elasticity of demand - people tend to buy a car when they need one.
3) Housing - AirBNB, speculators buying up houses, and collusion among apartment complex owners. Also the misforecasting of demand by manufacturers of lumber, plywood, etc.
4) Price gouging - certain industries have used the claims of high inflation to massively increase prices and profits even though the underlying factors should result in a modest increase. So we saw absurd prices for eggs and absurd profits.
5) Labor shortages due to massive retirement
The oil prices are largely collusion driven by OPEC.
So we have 3 areas without any real elastictiy of demand that have had shortages due to collusion and incompetence. An general increase in prices because claims of inflation give cover for collusive increase in prices for price gouging.
Then we have a shortage of labor due to retirements.
Neither the summary nor the article explain why OpenAI was interested in buying the company.
MineCraft has been one of the training grounds for OpenAI research, they can use this instead - it looks better, and they can run it drastically faster for training on their hardware.
ChatGPT hasn't "gotten worse".
It just switched the bias for whether it was saying a number is prime or not. Previously it was most likely to say yes, a test set that was rich in primes it would get lots of correct by accident. Now it is most likely to say no, which means a prime rich test set it will get mostly wrong.
It was never 'using math' to identify primes, only memorization.
#1 is a lot more than remotely plausible;
Bat coronaviruses aren't structurally adapted to humans, they need to infect an intermediate host to become more adapted to a physiology more closely to that of humans.
So a direct from bat to human for this coronavirus is incredibly unplausible. The most likely source is actually an infected animal from an exotic animal farm that did deliveries to the Wuhan wet market.
Waymo's do well in light rain and fog now. Don't know if they are letting them drive independently yet but if not they probably will soon.
Not sure when they will be ready for heavy rain, ice, and snow though.
No details about the model, only thing provided was that it has improved performance. Which is fairly 'duh'.
They are grouping by education. If the average education requirement for jobs has increased and thus people seek more education it makes each group 'dumber' on average.
So if the most intelligent of each group has now moved to the next highest education level, the average IQ of the group they leave drops, and the average IQ of the group they move into drops.
So this alone isn't evidence that the population is less intelligent, but rather is evidence of education requirements inflation.
It is a large language model, it isn't "ChatGPT" style. What makes ChatGPT different from other large language models is it was tuned for chat.
They needed to illustrate how good their T-800 without being too obvious. This was the most realistic non-lethal scenario they could come up with. They are a military application company, not a creative studio, you have to forgive them.
They haven't been a military application company for many years. First they were acquired by Google, then Hyundai Motor Group.
There is no path to get Atlas to any sort of actual application, it overheats fairly quickly, so can only be used for short demos, not real work.
Every website has a TOS. The AI researchers are violating it simply by downloading these images.
TOS only matter if you have to agree to them ('click license') to visit and see the images (such as a click license when you sign up for the site, or when you first visit the site, etc.). If not then they aren't actually a contract, and thus only normal copyright law applies.
The project will seek to integrate with existing open map data from projects such as OpenStreetMap and city planning departments, along with new map data contributed by members and built using computer vision and AI/ML techniques to create a living digital record of the physical world.
It's all about GPU processing power. Rent time on Vast.AI, use the free (or paid) tier on Collab, or throw a 3090/4090 or upcoming 4090 Ti at it. Anything less than a 3080 and you'll probably have a poor experience. CPU would be a nightmare.
Even a 3060 mobile variant is quite nice. The faster GPUs are nice to have, and essential for training your own model due to VRAM, but for inference a 3060 is adequate. Might change with larger models though.
Money is the root of all evil, and man needs roots.
