Adam Fourney

@[email protected]
256 Followers
56 Following
166 Posts
Principal Researcher; Human-AI eXperiences (HAX) Research Group; Microsoft Research
adamfourney.comhttps://www.adamfourney.com/
Microsoft Researchhttps://www.microsoft.com/en-us/research/people/adamfo/
Adam FourneyOct 21

I am extremely excited about this work! It opens a new and useful way of using agents (addressing a new category of "FOMO" tasks). And, it's highly generalizable: monitor prices, paper pubs, GitHub, Discord -- even traffic cameras, all with a simple text prompt

https://www.microsoft.com/en-us/research/blog/tell-me-when-building-agents-that-can-wait-monitor-and-act/

SentinelStep teaches agents to wait, monitor, and act.

SentinelStep enables AI agents to handle monitoring tasks that run for hours or days, like watching for emails or tracking prices. It works by managing when agents should check and their context, avoiding wasted resources and missed updates.

Microsoft Research
Show threadAdam FourneyMar 26, 2023
@jbigham A lot of excitement recently has been around what you can do with *fewer* parameters.
Show threadAdam FourneyFeb 28, 2023
@meg I enjoyed it. I would have liked him to spend a minute or two discussing a little more about how LLMs work — specifically how pre-training basically just turns the Internet into a series of madlibs. This makes it pretty easy to see why hallucinations might arise.
Adam FourneyFeb 27, 2023
@andresmh No kidding. Remember Heartbleed? World should have learned then how critical OSS infrastructure can be (in that case OpenSSL)
Show threadAdam FourneyFeb 19, 2023
@roguechi bah, don’t write off bm25 just yet… something has to feed those prompts
Adam FourneyFeb 18, 2023
@andresmh all 4 of those
Show threadAdam FourneyFeb 17, 2023
@grahamcox82 @JenMsft Figures from papers and presentations I was working on… OneDrive casts a wide net when indexing images…
Show threadAdam FourneyFeb 17, 2023
@ct_bergstrom pretty sure “guessing” is most accurate
Adam FourneyFeb 15, 2023
@andresmh I’m clearly very biased, but MSR had a ton of impact on Bing, for example. Just overwhelming (e.g Learning to rank) And, Bing + MSR has inspired other parts of the company to take a more data-driven, experimental approach to development. Heck, look at Jaime Teevan’s meteoric rise. But it’s not flashy. There are no press conferences. And product teams have started to build their own applied science teams to run those experiments. So there is a need to adapt I think.
Show threadAdam FourneyFeb 13, 2023
@Mor @katestarbird Peter remembers: https://twitter.com/peteratmsr/status/1621978644892762112?s=46&t=2OMc9Il1uB8uJ92MGGV4Kg
Peter Lee on Twitter

“For the record, I had nothing to do with the Chinese balloon despite my previous history of big balloons in the military... https://t.co/CUdILK6TeY #shotdown”

Twitter