So search engines in their traditional sense will be obsolete anyway.
1) GPT-4 and other such LLMs will generate textbooks and manuals for every conceivable topic.
2) These textbooks will be 'dehallucinated' and curated by known experts on particular topics, who have reputations to maintain. The experts' names will be advertised by the LLM provider.
3) People will search for stuff by chatting with the LLMs, which will in turn provide citations for the chat output from the curated textbooks.
There are weaker formal systems like Presburger arithmetic (peano without multiplication) and Skolem arithmetic (peano without addition) that have been proven to be complete and consistent. Tarski also showed that there are formal systems for real numbers (hence also geometry) which have the same properties. (although the real numbers include integers, the integers alone have a lot more structure and so Tarski's result does not imply Peano)
There are also extensions to these (eg. presburger extended to multiplication by constants) that are also known to be complete and consistent.
These systems do not require any social compact. Any theorems proven through them are absolute truth, although the the range of statements that these systems can express is limited.
One may require a social compact for Peano, ZFC, and such other powerful formal systems.
That the software implementations like Coq and Lean are bug free may also require a social compact, if the nature of being bug free cannot be formally proved, although it seems determining this should be an easier problem.
Rather than asserting that current LLMs are at their tail end, or that AI isnt good enough, it is much more instructive to check what are the bottlenecks or constraints to further progress, and what would help remove these bottlenecks.
They can largely be divided into 3 buckets
1) Compute constraint - Currently large companies using expensive nvidia chips do most of the heavylifting of training good models. Although chips will improve over time, and competition like Intel/AMD will bring down prices, this is a slow process. But what could be a faster breakthrough is training using distributed computing over millions of consumer GPUs. There are already efforts in that direction (eg. petals/swarm parallelism for finetuning/full training, but the eastern europe/russian guys developing them dont seem to have enough resources).
2) Data constraint - If you just rely on human generated text data, you will soon exhaust this resource (maybe GPT4 has already). But the Tinystories dataset generated from GPT4 shows if we can have SOTA models generate more data (and especially on niche topics that appear less frequently in human generated data), and have deterministic/AI filters to segregate the good and bad quality data thus generated, data quantity would not be an issue any longer. Also, multimodal data is expected (with the right model architectures) to be more efficient at training world grokking SOTA models than single modal data and here we have massive amounts of online video data to tap into.
3) Architectural knowledge constraint - This may be the most difficult of all, figuring out what is the next big scalable architecture after Transformers. Either we keep trying newer ideas (like the stanford hazy research group does), and hope something sticks, or we get SOTA models few years down the line to do this ideation part for us.
one (quite convincing) theory is that anything that can be achieved by a carbon-based neural network (eg. human brain) can also be achieved by a silicon-based neural network. The hardware may change, but the hardware's software expressiveness shouldnt be affected, unless there is a fundamental chemistry constraint.
Since human brains during dreams (lucid or otherwise) can generate coherent scenes, and transform individual elements in a scene, diffusion based models running on cpu/gpus should eventually be able to do the same.
How can a declining population selling mostly commoditized goods support ever increasing property prices, that too when property is already overleveraged!!
This crash was always on the cards. Just a matter of when, and the when may have finally arrived.
I get the feeling the article preaches to the choir.
The serious sources have always portrayed NIF's work as technical achievements. But they are read mostly by scientist and engineer types.
Mass media which hypes things is read, well by the masses, who dont have the patience or inclination to delve into technical details.
This dichotomy will always exist. I remember once reading a Chekov story where two intellectuals discuss how the townspeople are more interested in silly affairs and scandals rather than recognizing intellectual achievements.
I like the fact that Andrew is very transparent and honest about his experiments.
Although he didnt explain 1) how did the Fe impurities get localized into certain regions of the overall batch. 2) Were the impurities quantitatively enough in the magnetically susceptible shards to cause the half levitations he demonstrated.
Hope he keeps running and reporting some side experiments, instead of completely going back to his day job.
Lets not drag Japan into this. They have very high standards in research despite the few examples you gave here, some of which are not even science related.
As a widely accepted measure of this, you can read about the many nobel laureates from japan.
Anyways lets keep politics out of the discussion. It was not my intent anyway, in case you perceived it that way.
All what you said may be true, but ASML would still be foolish to trust and export to China, given that the latter is hell bent on copying and reverse engg its products, even if unsuccessfully for now.
Dont assist even your incompetent adversary. Competition 101.
Lets just say semiconductor manufacturing is the best example of regional competencies and global trade.
Globalization has failed on many fronts and resulted in race to the bottom outcomes, but here it works well due to specialization.
Having understood that, one can also see a core TSMC strength is that Taiwanese workers are disciplined and work long hours for moderate pay, a culture that European and American labor will never accept. So all these new factories in US/EU should be treated as high cost alternatives, justifiable only for non cost-sensitive customers like defence/aerospace/high end cars.
No,even though the 10 pikachu fans assess independently, their estimations will be correlated. As you may already know the overall estimate wont be 100%, rather about 65%. The formula when the assessors are known to be independent, is simpler and has already been indicated by another commenter, P(A1 or A2 or A3 or ..An) = 1 - P(not A1)P(not A2)..P(not An)
On the other hand, if those 10 were engaged in complete groupthink (so no additional information beyond the first guy), the overall estimate would remain 10%.
In general, the answer would lie between 0.1 and ~0.65 depending on how their estimates influence each other.
Alright lets do some calculations. Feel free to dissect the assumptions.
K event = Kim says Material L is RTSC
D event = DFT says L is RTSC
S event = L is actually RTSC
we have assumed P(S/K) and P(S/D) are each 0.1, though we could have chosen other numbers for them as well.
We want to estimate P = P(S/(K and D))
P = P((K and D)/S)P(S)/P(K and D)
Assuming Kim and DFT are in the business of making positive predictions, they always get it right when L is actually RTSC.
so P(K/S),P(D/S) and P((K and D)/S) are all taken as 1
hence P = P(S)/P(K and D) = P(S)/(P(K)P(D/K))
= P(S/K)/P(D/K) = 0.1/P(D/K)
(similarly, P = P(S/D)/P(K/D) = 0.1/P(K/D))
But ofcourse we dont know P(D/K) or P(K/D). We could check historical data on how often D aligns with K, a messy exercise at best. Say they dont perfectly align, then the conditionals are less than 1,and P>0.1. Even intuitively when K (or D) gets additional support in the form of D (or K), your P should increase, not decrease.
If we assume D and K align on average, then P(D/K) or P(K/D) is 0.5, and we get P = 0.2.
You may estimate everything above differently, thus getting a different P. You can also come up with your own way of modelling this. I came up with my particular estimate to understand how frequently should i follow the news, and care about the whole thing. You should model it according to your usecase.