Bill Howe

@nntaleb Secrets are news you tell one person at a time

I have a secret to share. If you want to spread information, make it look like a secret. But don't tell anybody about this. (Version 2)

@pmddomingos @geoffreyhinton Predictably ad hominem. That a politically motivated disemboweling of world-leading research funding apparatus might raise questions of scientific integrity is not an “embarrassing” position to take. But of course you know that.

@geoffreyhinton Don’t embarrass yourself, Geoff.

I think Elon Musk should be expelled from the British Royal Society. Not because he peddles conspiracy theories and makes Nazi salutes, but because of the huge damage he is doing to scientific institutions in the US. Now let's see if he really believes in free speech.

🚨Curious how LLMs deal with uncertainty? In our new #EMNLP2024 Findings paper, we dive deep into their ability to abstain from answering when given insufficient or incorrect context in science questions 💡arxiv.org/pdf/2404.12452 Joint work w/ @billghowe @lucyluwang @uw_ischool

@AdamKuhr @PeteButtigieg Another day another reason @petebuttigeg for president

This is @PeteButtigieg with the only acceptable answer.

@Eric_Erins This isn’t the response tho?

Big Tech is almost over we’re almost free

@GreatDismal evil yoga

Son texts me that he’s sending a link to an article about “demonic positions”.

@rachelmetz Agree, but I also fear the idea that because a paper *is* peer reviewed it earns free credibility. There exists unintentional gatekeeping, especially in contentious domains. Arxiv has been a positive. (Overlay journals are one underused tool in the toolbox.)

It’s Saturday, which means it’s time for a reminder that papers published to the arxiv are *not* peer reviewed and should be approached with caution (if not skepticism), particularly if you’re a journalist (or anyone interested in sharing research with a large audience).

@ylecun @scienceisstrat1 @altcap @Jason @chamath @DavidSacks It’s true that, like reality labs, none of those agencies generate direct revenue. But having trouble seeing how this comparison makes a favorable case for metaverse. It was a profit-motivated bad idea, sans any hope of profit.

@scienceisstrat1 @altcap @Jason @chamath @DavidSacks Does anybody understand what a long-term investment is? Would NSF, NIH, & DARPA be described as "money pits"? If "science is strategic," then what is long-term R&D if not strategic? Who can afford to invest in long-term transformative techs if not governments & tech companies?

Meta’s metaverse money pit Cc: @altcap @Jason @chamath @DavidSacks

@ylecun Querying gridded data <BLANK> <BLANK>

Learning representations with multiple layers.

@Shahules786 @ylecun Compare likelihood of surviving the day

@ylecun An average human on earth today works much more hours than our hunter-gatherer ancestors. And yet, can we say that we are happier than our ancestors?

History shows over and over that society and people's well-being makes progress with more intelligence: better skills, literacy, education, creativity, culture, communication, & the free exchange of ideas. Intelligence, by humans and machines, is what fuels progress.

@ylecun C’mon Yann. There are 100M active daily users of the plane in your analogy, and countless other mini-planes of the same design deployed in all kinds of contexts. For five years everyone has raised VC and their stock price by claiming to use a plane somewhere in their system.

Many AI safety discussions today seem as speculative as discussions about airliner safety in 1890. Before we have a basic design & basic demos of AI systems that could credibly reach human-level intelligence, arguments about their risks & safety mechanisms are premature.

@ylecun While simultaneously demonstrating the sexual objectification bias, to the extent that women can literally be rendered as a collection of body parts. Work conceived and led by @wolferobert3 and @aylin_cim with some support from myself and @yanyiwei_44 arxiv.org/abs/2212.11261

Teratogenerative AI : producing monsters with AI.

@arokem @uwresponsibleai ChatGPT’s answer sounds about right; we made it up. We didn’t want talking heads for a topic so pervasive in public discourse.

@uwresponsibleai What's a "reverse panel"?

Our first RAISE Seminar this Friday, 1/13, at 9:00 am will be a “reverse panel” discussion: ChatGPT and the Future of Information Work. RSVP and submit your questions about ChatGPT here: forms.gle/YUNVffTW6yh5Wr… Zoom link: washington.zoom.us/j/96044346127 #responsibility #AI

@stephmui @triketora almost?

The biggest difference I didn’t expect about being a VC vs founder was the mindset shift from being skeptical to being almost delusionally optimistic. In VC, your job is to say no 98% of the time. As a founder, your job is to sustain belief while being rejected 98% of the time.

@nathaliejacoby1 of denial

@ylecun Sand isn’t as plentiful as it seems. newscientist.com/article/231317….

Facts: nuclear is cheap & zero emission, wastes can be stored safely, nuclear accidents have killed very few people. Solar panels are made of silicon, basically sand, which is plentiful. They are made in sophisticated factories by highly educated workers and engineers.

- tweet: pic of a nuclear power plant - comments: so much CO2 emissions! whaddabout wastes? Too expensive! Remember Fukushima & Chernobyl? - tweet: pic of a large solar array in a desert - comments: what a waste of space! All that rare earth comes from China! Child labor! ....

@amyjko And now you can compare the traffic before and after the change to see which maximizes engagement? Wait, never mind.

In my endless quest to resist academic bean counting, I replaced the projects list on my website (which listed people, papers, grants, etc.) with a list of discoveries, which aggregate multiple papers into claims: faculty.washington.edu/ajko Things I learned while writing them 1/n

@domoritz @random_walker @mmitchell_ai Well it’s not strict. The main thing is that the question of “does it actually work” tends to be almost an afterthought in CS, and I want to see that story before I slog through 6 pages of pedantry. Related: I try to understand figures first, reading just enough to do so.

@billghowe @random_walker @mmitchell_ai Sigmod papers having the related work in the end kinda messes that up. I prefer having the related work first, though, to get the context.

15 years ago my PhD advisor taught me One Weird Trick for editing your own writing. Edit **back to front**, paragraph by paragraph. I still use it and it still surprises me how well it works. When I get my students to do it, it often blows their minds. Try it!

This is a major milestone for US AI Ethics law. Brings in groundwork for: - Safety - Equity - Data Rights - Transparency - Consent Yay for @SureshVenkat46 and @kdphd 's work I know must underlie this! technologyreview.com/2022/10/04/106… Thanks for the reporting @Melissahei