It's this fetishization of the seemingly more "rational" position that these people seem to have. It's similarly mind-boggling when people have discussions about free-will and determinism. Free will is quite possibly the most fundamental experience humans have yet people will flatly deny free will exists.
>Similar to the Chinese room argument, one can always claim the computer is just emulating understanding.
And you'd be correct. The point isn't what kind of output ChatGPT can produce, the point is what kind of input it takes to create the model.
If ChatGPT were to gain language understanding at the level of a toddler trained on the same number of tokens that a toddler needs, then you could start postulating that the model is really learning, rather than becoming a sophisticated stochastic parrot.
The NSA is the biggest of Brothers in the world and supposedly the country keeps on innovating and big tech power is strongest here. How do you square that circle?
One thing people generally don't highlight about the transformer architecture, which I think is very important, are the pass through connections. This allows the model to "spread" out it's learnings across the network rather than them being forced to very local. I think this is also a reason why the Resnet was so successful.
I never understood newtonian mechanics taught in high school physics until I wrote a simulator for it. The code is generally ~100 lines in python but you really get the feel for how the equations work.
It's not very surprising though. Most of Silicon Valley and HN royalty is staunchly libertarian. VCs doing zero labour and getting rich off the backs of workers is a phenomena that cant easily be defended unless you buy into this ideology.
Maintenance of the power inverter, batteries, wiring, and the panels themselves means that marginal cost is not zero, and as long as we don't know the solution to entropy, will never be "essentially" zero.
Exactly my experience trying to get GPT-3 to produce SQLITE queries. Anything with a CTE, it gets hopelessly wrong. I was super excited to show off this to my team and after one too many false results I just quietly swept this idea under the rug.
How is it weird? The point is to compare to a “safe” investment. A successful startup is supposed to have 10x or 100x returns for investors. A 30% return should be considered a failure.
Deepmind is high status as well. However with both these companies - they are looking for experienced researchers usually rather than new grads. Google was famous for hiring bright eyed young geniuses and building their careers.
This perception changed 5-7 years ago. I actually can't think of a single software company that induces the same level of awe in university students anymore. Places like Jane Street do but they aren't really Silicon Valley tech companies.
It's interesting reading all these comments reaching for concepts not quite in peoples vocabularies because they have been stripped out systematically. Its cruel, inhumane, how can they just make a unilateral decision about something I devoted myself to, companies are making billions in profits etc. Marx wrote about all this 150 years ago in great detail.
You are being sarcastic - but this mindset change should happen in American cities if we are to move towards a sustainable future. We should have regular roads for bicyclists and pedestrians and "car lanes" on some roads, "truck lanes" on others etc.
It's not a traditional FPGA chip (lots of luts and flip flops). The "AI Engine" is basically hardened chiplets that are working alongside soft logic chiplets and I/O. This is how they're able to get their performance/power numbers
It's really dangerous for AIs and their proponents, actually. Once people start mistrusting these results, it'll be a slippery slope. I've generally just stopped using ChatGPT entirely because of this.
You don't even need to retrain if you use retrieval transformers. That is the real revolution waiting to happen. Deepmind already unlocked it with RETRO, but I don't know why a public version hasn't been released - hooked into the live internet.
If anything that's what makes them more dangerous. I tried Copilot for a month and ended my subscription. It will write a SQL query with all the expected keywords in there: a WHERE for a filter, an ORDER BY for a sort, but the way it combines these things, especially recursive queries, is almost always wrong except the most trivial queries which I would have written by hand in no time anyway.
You are misunderstanding the situation. Karpathy claimed that sensor fusion of radar, sonar and vision isn’t working well. He made no such claim about Lidar. Lidar is the sensor that is the crucial difference between Waymo and Tesla's approach to self driving.