It's pretty easy to make $100K per year doing consulting at half-time or less, and have 50% or more time left-over to do whatever you want. If there is any argument about staying in academia, it cannot be rooted in financial considerations.
To play devil's advocate, the value these prestigious journals provide is providing an objective function to optimize allocation of resources in academia. To advance one's career in academia, one needs to publish in prestigious journals, there isn't really a way around it (and if the entire system were to change, resources would need to be invested to create a different hierarchy of publication quality). To whatever extent publishing in prestigious journals captures scientific quality (and/or political astuteness), it serves the purpose of funneling more resources to those that optimize that objective. So it's not really a scam when everyone has run their individual game theory and chosen to continue with this particular system, and these journals can extract their cut because there is still value in publishing in Nature/Science/Annals or whatever. There is a push to change some of this but typically only the most successful academics can afford to boycott these journals, after already having dozens of publications vetted by the existing system.
The model probably takes into account the state as a variable, and states vary as far as average hesitancy goes and in terms of vaccine distribution strategy.
Unlikely due to the buyers also being sophisticated. "Sequoia and Steadfast Financial bought shares at a $5 billion valuation from some of Zapier’s original investors."
It's strictly speaking correct in that the $5B valuation was established purely on the secondary market and only from sales of shares of investors (not common holders), which doesn't give the company any VC funding. Also $1.3M in seed funding is a rounding error to get to $5B.
Not advocating for the pool idea but that reasoning is generally incorrect. Decreasing personal capital/timing risk actually increases founder risk tolerance for bigger outcomes for their startup. Investors should want founders to be hungry for big capital, not small.
Impressive to a human is a highly subjective property. Humans generally consider the understanding of language to be an intelligent trait, yet tend to take basic vision which took much longer evolutionarily to develop for granted. Neural networks can approximate arbitrary functions, and the ability to efficiently optimize neural network parameters over high dimensional non-convex landscapes has been well established for years. What typically limits pushing the state of the art is the availability of "labeled" data and the finances required for very large scale trainings. With NLP, there are huge datasets available which are effectively in the form of supervised data, since humans have 1) invented a meaningful and descriptive language and 2) generated hundreds of trillions of words in the form of coherent sentences and storylines. The task of predicting a missing word is a well-defined supervised task for which then there is effectively infinite "labeled" data. Couple these facts with a large amount of compute credits and the right architecture and you get GPT3. The results are really cool but in my opinion scientifically unsurprising. GPT3 is effectively an example of just how far we can currently push supervised deep learning, and even if we could get truly human level language understanding asymptotically with this method, it may not get us much closer to AGI, if only because not every application will have this much data available, certainly not in a neatly packaged supervised representation like language (such as computer vision). While approaches like GPT3 will continue to improve the state of the art and teach us new things by essentially treating NLP or other problems as an "overdetermined" system of equations, these approaches are subject to diminishing returns and the path to AGI may well require cracking that human ability to create and learn with a vastly better sample complexity, effectively operating in a completely different "under-sampled" regime.
"The mistake SoftBank made wasn’t investing in them, it was investing at a valuation that grossly overestimated the size of the opportunity, and how quickly they would see their return."
That doesn't say much. When you're off by 20X for a close-to-IPO company that's a huge mistake and the diff in funds would've been better spent on spray and pray R&D. Many companies are good investments at 20X lower valuations than what they recently raised at.
If you're doing something for which the opportunity cost of waiting two years is not very high in the first place, perhaps it's not worth doing at all.
Pretty much my experience as well moving from Russia to the US. Team sports and social deviance seem to dominate the popularity hierarchies of American schools (as of 2003). In Russia being the top student is respected instead.
"I think on average the difference between a high school student from Russia and US is negligible." That's difficult to believe. I went to comparable level early education schools in Russia and the US and observed stark differences in the quality of math education. Math culture is a thing, and it varies between countries, affecting the education system on the whole as well.
"we're definetly at least 15+ years out". Similar statements were made about Go the year it was solved. AV is a vastly harder problem and requires new techniques to get there, but AI can progress can happen any time.
Presenting a subset of the information to let the uninformed jump to favorable conclusions for the presenter is not a new marketing strategy. If there's no indication about the true level of progress, what is the purpose of the demo?
This demo is not informative as to the readiness for scalable L4 deployment, for which it would be necessary to focus on the breadth/accuracy of perception features under the hood of intent prediction and what happens at the tail end with arbitrary situations that occur in urban driving environments.
Quite a nice result. Along the lines of your point 2., it's not the first time in recent years a problem in operator algebras was resolved by the computer science community using unexpected methods: https://www.quantamagazine.org/computer-scientists-solve-kad.... Perhaps looking at these problems from a more concrete angle is the way to go, especially for a problem like disproving the Connes conjecture.
"the state-of-the-art in computer vision has moved so significantly that it’s arguably now not the primary blocker to commercial deployment of self-driving cars."
The state of the art in machine perception is still far from human level perception, regardless of how quickly progress has been made. Prediction is a function of perception, so it's also bottlenecked on perception getting better. Let's not confuse recruitment-oriented blogposts with credible scientific writing.
They would have able to win and get away with it if they incorporated the knowledge of the external dataset directly into the ML model, provided they had a reasonable estimate on the fraction of overlap between the external data and the test set. A weak version of this would be to just train on the external data in addition to the provided data. A stronger version would train regularly on the provided training data and in addition overfit on a random subset of some percentage of the external data (with some small random prediction error thrown in to obfuscate), which would get equivalent results to what they did with logic.
It takes a while to internalize what's difficult about the problem. Partly I think it's that humans take for granted their ability to see, which requires no conscious effort but expends 70% of the brain's computational resources at any time. It's not clear AGI is required... computer vision is not known to definitely be AI-complete. It's a very difficult problem but could potentially be solved in the next several years.
It's simple: choosing again switches from a losing outcome to a winning outcome and vice versa, thus the probability of winning by choosing again is 2/3.
computer vision is clearly much harder than go. the point is that the forefront of AI is moving quite rapidly, and the consequence are bound to catch the casual observer (or even an expert) off guard.