A conversation about AI for science with Jason Pruet
1–10 of 163 posts
Re: A conversation about AI for science with Jason Pruet
#2PR in here for sure, but some smart context on the scientific and nat security potentional the DOE and National Labs see in AI.
Re: A conversation about AI for science with Jason Pruet
#3wonder if they still train all of their models using Mathematica because it was impossible to get pytorch on the classified systems
Re: A conversation about AI for science with Jason Pruet
#4The real title is "Q&A with Jason Pruet"
Re: A conversation about AI for science with Jason Pruet
#5wonder if they still train all of their models using Mathematica because it was impossible to get pytorch on the classified systems
AFAIK that was mostly due to a silly detail about MD5 hashing being restricted on FIPS compliant systems? Or something like that. I'm pretty sure there's an easy workaround(s).
Re: A conversation about AI for science with Jason Pruet
#6> We certainly need to partner with industry. Because they are so far ahead and are making such giant investments, that is the only possible path.
And therein lies the risk: research labs may become wholly dependent on companies whose agendas are fundamentally commercial. In exchange for access to compute and frontier models, labs may cede control over data, methods, and IP—letting private firms quietly extract value from publicly funded research. What begins as partnership can end in capture.
Re: A conversation about AI for science with Jason Pruet
#7wonder if they still train all of their models using Mathematica because it was impossible to get pytorch on the classified systems
AFAIK that was mostly due to a silly detail about MD5 hashing being restricted on FIPS compliant systems? Or something like that. I'm pretty sure there's an easy workaround(s).
there were a bunch of reasons. couldn’t bring compiled binaries onto the red, so you had to bring the source + all deps onto a machine with no external internet.
it was unpleasant.
Re: A conversation about AI for science with Jason Pruet
#8I was a bit puzzled what "1663" is. Here's what I found:
> The Lab's science and technology digital magazine presents the most significant research initiatives and accomplishments from national-security-related programs as well as projects that advance the frontiers of basic science. Our name is an homage to the Lab's historic role in the nation's service: During World War II, all that the outside world knew of the top-secret laboratory was the mailing address - P.O. Box 1663, Santa Fe, New Mexico.
https://researchlibrary.lanl.gov/about-the-library/publicati...
Re: A conversation about AI for science with Jason Pruet
#9I like how he says that AI is a general-purpose technology like electricity.
Re: A conversation about AI for science with Jason Pruet
#10> We certainly need to partner with industry. Because they are so far ahead and are making such giant investments, that is the only possible path. And therein lies the risk: research labs may become wholly dependent on companies whose agendas are fundamentally commercial. In exchange for access to compute and frontier models, labs may cede control over data, methods, and IP—letting private firms quietly extract value…
WILL end in capture. Profit demands it.