I fear a lot of AI benchmarks, papers, studies, surveys and such are little more than PR, lobbying and propaganda at this point. There are too many lucrative financial interests now that more outrageous claims are being made by the month. If the 2017 paper was the beginning, I believe by 2027 we might have a more realistic view of the actual value of LLMs and their likely trajectory.
It's also possible that there are many other better architectures for AI to pursue that make up for many of the blind spots, limitations, hallucinations and scaling bottlenecks of LLMs yet to be discovered or commercialized.
Yes, I agree completely about all the hype. That being said, using the technology that exists today, I think it’s possible to build a research assistant that can automate much but not all of the work needed to review some types of academic papers. That AI app won’t replace people on net, but it will make them much more productive,
I fear a lot of AI benchmarks, papers, studies, surveys and such are little more than PR, lobbying and propaganda at this point. There are too many lucrative financial interests now that more outrageous claims are being made by the month. If the 2017 paper was the beginning, I believe by 2027 we might have a more realistic view of the actual value of LLMs and their likely trajectory.
It's also possible that there are many other better architectures for AI to pursue that make up for many of the blind spots, limitations, hallucinations and scaling bottlenecks of LLMs yet to be discovered or commercialized.
Yes, I agree completely about all the hype. That being said, using the technology that exists today, I think it’s possible to build a research assistant that can automate much but not all of the work needed to review some types of academic papers. That AI app won’t replace people on net, but it will make them much more productive,