Zeve Sanderson

Angeleno trying to understand snow and social media

Executive Director @csmap_nyu by day, dissertating @VUamsterdam by night

[Zeve Sanderson @ZeveSanderson]

my mom has been a graphic designer for 40 years, and today when i showed her DALL-E and explained how GPT-3 ~works, she was deeply saddened that the very artists this tool will displace created the 'training data' that enabled this system to be built in the first place

[Centro de Estudios en Libertad de Expresión (CELE) @CELEUP]

📄Nuevo paper de @JoanBarata , en colaboración con @CELEUP, sobre #GenerativeAI, publicidad y regulación 🧵👇

[Zeve Sanderson @ZeveSanderson]

To Zeynep’s always astute reflection, AI itself seems unconvinced by the methods used in or conclusions asserted by the article

[zeynep tufekci @zeynep]

Such quizzes are fun, but this doesn’t get at anything interesting or consequential about AI beyond the most important thing about all this: AI can produce human language.

Taking this quiz too seriously, though, may well be a good indicator of missing that actual key point.

[zeynep tufekci @zeynep]

Such quizzes are fun, but this doesn’t get at anything interesting or consequential about AI beyond the most important thing about all this: AI can produce human language.

Taking this quiz too seriously, though, may well be a good indicator of missing that actual key point.

[James Bisbee @JamesBisbee]

1/12
As conversational AI replaces search, we asked a simple question:
Do chatbots present the same political reality to everyone?
Short answer: no.
In a new audit of ChatGPT and Grok, we find systematic ideological adaptation in multi-turn conversations. 🧵

[Francesca Whitehead @FLFWhitehead]

@AnjneyMidha Strong agree. This is why I left investing to work on this full-time. We’re building enterprise context management systems at Pacific, starting with financial services, where information governance, portability, and reliability are mission-critical.
https://t.co/EfqCxxWfTZ

[Pacific @pacificint_]

To outrun their competitors with AI that cannot only leverage raw intelligence, but the information that makes their company their company. And do so in a monitored, permissioned and secure way that leads the industry on speed and quality.

Just as AI foundation labs put raw intelligence an API call or query away, Pacific does the same for your internal data.

[Ben Scharfstein @benscharfstein]

Context is the product (see quote tweet)!

Couldn’t agree more with the thesis here and @noahfradin and team are absolutely incredible

[Jeff Morris Jr. @jmj]

We went from mainframes to personal computers. Now it’s time to go from generic AI to personal AI.

Excited to share a stealth investment from last year - introducing @pacificint_, building the context layer for AI.

We led the deal with @zoink. This team is special.

[Francesca Whitehead @FLFWhitehead]

Most AI still treats context as an afterthought.

Teams copy info into prompts, re-upload files across tools, build one-off workflows, and rebuild context as they switch systems.

@pacificint_ changes that.

[Zeve Sanderson @ZeveSanderson]

Our software through our data and context. When software can be generated on the fly, it's the information that it acts on that is unique to you, and that makes it work for you."

Check out the blog post and follow the company. Excited to see where they can take it!

[Zeve Sanderson @ZeveSanderson]

Such a great team to be working on such an important problem!

As everyone in the space knows, context management is a key missing layer in unlocking the potential of ever-more-powerful models for the use cases that matter most across the workplace and home.

[Pacific @pacificint_]

Every company and every person will one day have a magical library that personalizes their computing experience. We're the team building it.

Meet Pacific:

[Zeve Sanderson @ZeveSanderson]

I love their framing around the history of personal computing, from mainframes to personal machines to SAAS to AI: "We are moving from a world where we personalize our computers by downloading personalized applications that suit our workflows, to a world where we personalize

[Pacific @pacificint_]

Every company and every person will one day have a magical library that personalizes their computing experience. We're the team building it.

Meet Pacific:

[Sam Koppelman @SammyKoppelman]

New, batshit insane story from @hntrbrkmedia — and a special pod with @PabloTorre — dropping tomorrow

[Noema Magazine @NoemaMag]

“We [must] figure out how our contemporary communication system can strengthen democracy, rather than myopically focus on correcting falsehoods.”

—@ZeveSanderson & @jsbrennen

https://t.co/olfpT6AYho

[Megan A. Brown @m_dot_brown]

Thrilled to have a new article published in
@BigDataSoc! 🥳🎉📊

With scraping becoming a more common data collection strategy for internet researchers, we cover the legal, ethical, institutional, and scientific ramifications researchers should consider. https://t.co/CCvNEiEPKN

[Zeve Sanderson @ZeveSanderson]

@mmitchell_ai @boazbaraktcs it also emphasizes that truth-seeking is a process w/ principles, not simply a reflection of facts. hao's transparent explanation of her process, engagement with criticism, and commitment to correction is the more useful signal than the original error IMO

[Zeve Sanderson @ZeveSanderson]

@mmitchell_ai @boazbaraktcs +1 -- i often conceptualize the dimension of interest as "people who correct errors" vs "people who don't." it's a more useful heuristic than "people who get things wrong" (everyone) vs "those who don't" (nobody)

[Noema Magazine @NoemaMag]

“The last decade has made clear that we aren’t going to fact-check or inoculate our way toward a healthier civic culture.”

—@ZeveSanderson & @jsbrennen

https://t.co/olfpT6AYho

[McCourt School @McCourtSchool]

How do we protect children in the age of AI? @ZeveSanderson, a Tech & Public Policy visiting fellow and executive director of @CSMaP_NYU, explores the growing gap between regulatory frameworks and the chatbots that children use daily. @techpolicypress https://t.co/SwnlNxYHGA