Eric Horvitz

Chief Scientific Officer, Microsoft

[Eric Horvitz @erichorvitz]

Patrick Winston’s last lecture on “How to Speak.”

In 2019, as I was getting mic’d to give a lecture at MIT, Patrick slowly came down the auditorium steps with a cane. He handed me a copy of his new book, and we caught up briefly.

My host later told me he was very ill and had made a special effort to come by and say hello. Several months later, he was gone. I hadn’t realized he had come to say goodbye.

I deeply enjoyed our collaborations over the years. A brilliant and wonderful man.
https://t.co/nu36p5vS5n

[Eric Horvitz @erichorvitz]

At Build, we announced the general availability of Microsoft Discovery, a platform for accelerating scientific and engineering R&D. Discovery orchestrates leading-edge scientific models through cycles at the heart of science: evidence → hypotheses → analysis & experimentation → refined ideas—with human judgment at the center. https://t.co/tsSuHybW7i @Microsoft @MSFTResearch

[Eric Horvitz @erichorvitz]

AI and the firm: Empowering organizations to climb the gradient to their own AI frontiers. @satyanadella

[Satya Nadella @satyanadella]

With the new MAI models and Frontier Tuning capabilities we announced today, we're focused on helping every company move from just consuming a frontier model to fully participating at the frontier.

[Eric Horvitz @erichorvitz]

@MayoClinic and @Microsoft announce a collaboration to develop AI models for health care.

Excited to work with colleagues at Mayo, a world center of excellence in patient-centered problem solving, to explore how AI can support people facing some of the most difficult diagnostic and therapeutic challenges.
https://t.co/WOAf9PdG1W
@satyanadella @GFarrugiaMD @MicrosoftBuild

[Eric Horvitz @erichorvitz]

On related work & directions with agents: opportunities and prototypes described in this video are still as fresh as they were decades ago. Enjoy! https://t.co/3wWMk3YRot

[Eric Horvitz @erichorvitz]

Seems like yesterday!

[Eric Horvitz @erichorvitz]

@DimitrisPapail The ancient Greeks would think nothing of converting shadows like that into a time—or the radius of the earth.

[Eric Horvitz @erichorvitz]

@rao2z Great to see you in person!

[Eric Horvitz @erichorvitz]

Seems like yesterday!

[Udara @TGUPJ]

when an agent should act on its own, when it should ask, when it should defer, when it should explain itself...

@erichorvitz in 1999

[Eric Horvitz @erichorvitz]

Exciting results & direction with learning from signals from the environment—with implications for continual learning about the world.
@DimitrisPapail @MSFTResearch

[Dimitris Papailiopoulos @DimitrisPapail]

https://t.co/n10GwfKYuY

[Zachary Horvitz @zachary_horvitz]

Think your model is safe? Ask it again!

Check out our latest paper on estimating the tail risks of LLMs... 👀

[Eric Horvitz @erichorvitz]

New results on the performance AI on the reasoning tasks of a physician | Science https://t.co/nztmvsZbWK

More here: https://t.co/wmvAweHLrA

[Eric Horvitz @erichorvitz]

A huge loss.

[nature @Nature]

Geneticist J. Craig Venter, best known for his role in sequencing the human genome, has died aged 79.

He spoke to Nature in 2023 about AI, sequencing the ocean – and why he had no plans to stop working.

https://t.co/FYq5jbMpxa

[Eric Horvitz @erichorvitz]

@yudapearl @burkaru @eliasbareinboim Shifting to email for discussion

[Eric Horvitz @erichorvitz]

:-) More so to say that causal inference machinery could be leveraged formally in weaving together—and also guiding evidence creation (at the right time and place)—to build compelling stories that compete with grounded causal models. Then again, people have been persuaded over centuries by propaganda composed on the back of an envelope.

[Eric Horvitz @erichorvitz]

@yudapearl @burkaru @eliasbareinboim Feel free to use a different word than "persuasion." I remain concerned about the feasibility of adversarial generative explanations--even if the "causal models," built from toothpicks, only stand for a short time.

[Eric Horvitz @erichorvitz]

@yudapearl @burkaru @eliasbareinboim Adversarial generative explanations know the models and understand what events to generate in the world to tip the scales to an alternate coherent story.

[Eric Horvitz @erichorvitz]

@julia_kiseleva @besanushi @martinagvilas @vidhisha_b Lovely! Wish I could be there this year.

[Stanford Department of Medicine @StanfordDeptMed]

When physicians & AI reason independently, & then combine their perspectives, diagnostic accuracy significantly improves, according to a new @npjDigitalMed study by Selin Everett, with senior authors @jonc101x, @erichorvitz & others. https://t.co/j7haVbvPcc