Profile avatar
hershgupta.com
Lead Applied Scientist, Responsible AI @BCGX | @bostonu.bsky.social alum | Data, AI, and strategy enthusiast | Open-source contributor Opinions are my own #bikeboston #coys 📍DC -> BOS
112 posts 75 followers 149 following
Regular Contributor
Active Commenter
Posts 23 Comments 27

Not good. Thank you to @wired.com for documenting and to the researchers for speaking up.

submitted 45 days ago • 0 comments

I feel like my ability to figure out whether someone knows their stuff when they're talking about AI has become so sharply honed the more time I spend switching contexts between technical and non-technical people

submitted 48 days ago • 1 comment

Anthropic’s MCP should be more widely used in enterprise

submitted 52 days ago • 0 comments

An X engineer posted this output from Grok to demonstrate how "good" their LLM is (CW: racism)

submitted 63 days ago • 1 comment

This is sort of the skeleton key to understanding a lot of shit: there 100% is a non-trivial amount of inefficiency and waste in the federal bureaucracy, and most of it is a *direct result* of ‘reforms’ that are meant to assuage the people who complain about “waste and inefficiency in govt”

submitted 77 days ago • 42 comments

The most effective responsible generative AI technique is to proactively gatekeep LLMs from decision-makers who don't know how transformers work

submitted 77 days ago • 1 comment

A lot of people don't know how useful Model Context Protocol is - easily Anthropic's most underrated feature. Allows Claude to use your computer and applications with natural language. Official and unofficial MCP servers, which are like extensions, here: www.mcpservers.ai

submitted 79 days ago • 0 comments

If you have to prompt engineer confidence scores, then you already know they're not going to be reliable. arxiv.org/abs/2412.14737

submitted 79 days ago • 1 comment

If you, like me, have a feeling of despair reading this news, I'd ask that you consider donating to the organizations who do this work: Global Fund to Fight AIDS, Tuberculosis, and Malaria: act.unfoundation.org/FJvB3vUCJUep... International Medical Corps: internationalmedicalcorps.org?form=Main

submitted 86 days ago • 0 comments

It's actually theft of taxpayers money when websites go dark. We paid for that data

submitted 87 days ago • 229 comments

okay so what legal authority does elon musk have to commandeer the office of personnel management and access reams of sensitive data? www.reuters.com/world/us/mus...

submitted 87 days ago • 919 comments

Today, we are publishing the first-ever International AI Safety Report, backed by 30 countries and the OECD, UN, and EU. It summarises the state of the science on AI capabilities and risks, and how to mitigate those risks. 🧵 Full Report: assets.publishing.service.gov.uk/media/679a0c... 1/21

submitted 90 days ago • 8 comments

positron is the best IDE for data science and it's been my daily driver for over a year

submitted 89 days ago • 1 comment

Another takeaway is that we've been Stockholm Syndrome'd to think pandas has a reasonable and intuitive API for data analysis

submitted 89 days ago • 1 comment

middle managers who've never written a single line of code or built an ml model before

submitted 93 days ago • 0 comments

Every time I visit the North End here in Boston, I'm grateful to not be swallowed up by a flood of liquid sugarcane extract

submitted 94 days ago • 0 comments

A strange disconnect I've been thinking about: in the 1940s, 2.5m Indian soldiers fought against fascism in WW2. Many of their children came to America in the 90s tech wave. Now some of their kids and grandkids support the same ideologies they fought against

submitted 95 days ago • 1 comment

#Boston friends: want to see how the city's services are measuring up? I built CityUpdate Boston - daily plain-language summaries of municipal performance data, from EMS response times to sign repairs. Check it out: hersh-gupta.github.io/cityupdate

submitted 96 days ago • 0 comments

I gave deepseek-r1 (q8_0) a math problem and it got there after 10 minutes of non-stop trial and error

submitted 97 days ago • 1 comment

The forceful intrusion of Copilot, OneDrive, etc from Microsoft at the OS-level all but guarantees that I'll be using Linux from here on out

submitted 102 days ago • 1 comment

Researchers: this is _not_ how you evaluate LLMs www.nature.com/articles/s41...

submitted 102 days ago • 0 comments

Public sector orgs with large compute resources keep getting suckered into cloud data contracts because nobody on staff knows what kubernetes is - what a waste of taxpayer money

submitted 102 days ago • 0 comments

To deal with cobras, the British Raj placed a bounty on them, which worked at first, but resulted in more cobras as people bred them to collect the bounty The US bans TikTok for China ties, so users flee to Xaiohongshu "little red book" which sounds a lot like Mao's Little Red Book

submitted 102 days ago • 1 comment