impressive to me, but sadly i feel a little misleading since this is only the data-science part of life sciences.
every few weeks though i test claude and chatgpt on their scientific reasoning and it has definitely improved over time. in my experience without specific instruction on what is known/unknown they typically are lagging behind the leading edge of the field (dev bio/pluripotency in my case). probably because scientific research articles are not open-source so they can't crawl them.
claude has definitely outperformed chatgpt in this regard however, it's scientific reasoning is impressive.
every few weeks though i test claude and chatgpt on their scientific reasoning and it has definitely improved over time. in my experience without specific instruction on what is known/unknown they typically are lagging behind the leading edge of the field (dev bio/pluripotency in my case). probably because scientific research articles are not open-source so they can't crawl them.
claude has definitely outperformed chatgpt in this regard however, it's scientific reasoning is impressive.