Matt Jensen

[Matt Jensen @MattJensenSea]

Cup of java in my hand, so warm,
A sunny, pinkish glow proclaims a morn.
Across a Sound, Olympic pyramids
Command a land from shiny, snowy lids.

[Matt Jensen @MattJensenSea]

@TheDeuce @CastielsGranny @TrungTPhan Meat was often horrible then, and sauces were often needed to make it edible. Even many sauces would go rancid, in their opaque bottles. Heinz's clear glass bottle, which earned customer trust, was itself a revolution in consumer goods.

[Matt Jensen @MattJensenSea]

@MichaelWarbur17 It was still common in 1960 to pop into a theatre at any time. Watch the second reel, and stick around to see the first reel. (Expression: "this is where I came in".) Hitchcock insisted people come at showtime and see it start to end.

[Matt Jensen @MattJensenSea]

@MichaelWarbur17 @jbouie @RadioFreeTom Behind the scenes shot, Hunt for Red October. h/t @MichaelWarbur17

[Matt Jensen @MattJensenSea]

@TrungTPhan Useful factors, but a question here: "The transistor radio exploded in popularity... millions of teens — home for the holidays — played the newly released singles from The Beatles on repeat." On their radios? 🤔

[Matt Jensen @MattJensenSea]

@FroTrades @harrylitman The judges' "We disagree" is about speculation, not about the customers being "would-be". The judges are saying it's not speculation, and not "guesswork" to say they will eventually get a gay marriage request. They're saying there is no reason to assume that will not happen.

[Matt Jensen @MattJensenSea]

@RadioFreeTom @johnwparenti @ellenbraaten There's an episode of Room 222 where two students start an activism project to reduce smog.

[Matt Jensen @MattJensenSea]

@SamGrundman @100YearsAgoNews Same, which is why I'm confused. It's like saying my stolen Toyota was worth a billion dollars, because that's how much the Toyota factory cost.

[Matt Jensen @MattJensenSea]

@thinkdefence Origami kayak

[Matt Jensen @MattJensenSea]

@ESYudkowsky @dschwarz26 's scenario is probably more plausible (& nearly present & concrete) to most people than Skynet switching itself on. If you want people to notice Skynet risk of 10 years, IMHO don't mock warnings of milder hell in 3 years; it doesn't preclude Skynet birth after.

[Hayden @the_transit_guy]

“America is too big for rail” and yet we literally had it. Almost everything we advocate for is a restoration.

[Matt Jensen @MattJensenSea]

@MrSahilLihas @AziziShekoofeh @Mysiak @alan_karthi @vivnat @thekaransinghal @Google @GoogleAI Seen TXGNN? Pretty recent work, combines many sources into large knowledge graph for disease->drug mapping. Claims 49% better accuracy than other methods. Their preprint: https://t.co/igY3P5lMjF

[Matt Jensen @MattJensenSea]

@HolstDoug @MuellerSheWrote https://t.co/q8kNhMp1qQ.

[Jim Fan @DrJimFan]

Reading @MetaAI's Segment-Anything, and I believe today is one of the "GPT-3 moments" in computer vision. It has learned the *general* concept of what an "object" is, even for unknown objects, unfamiliar scenes (e.g. underwater & cell microscopy), and ambiguous cases.

I still can't believe both the model and data (11M images, 1B masks) are OPEN-sourced. Wow.😮

What's the secret sauce? Just follow the foundation model mindset:
1. A very simple but scalable architecture that takes multimodal prompts: text, key points, bounding boxes.
2. Intuitive human annotation pipeline that goes hand-in-hand with the model design.
3. A data flywheel that allows the model to bootstrap itself to tons of unlabeled images.

IMHO, Segment-Anything has done everything right.

[Matt Jensen @MattJensenSea]

@Scobleizer @fchollet @KeyTryer Can we swing from newness factor and ask what could be good analogies for the *scale* of disruption? "Is AI just disruptive to industries, or an existential threat?" There's Guttenberg, & Internet, but AI's growing exp., & may have potential for agency/intentionality; that's new.

[Matt Jensen @MattJensenSea]

@Scobleizer @fchollet @KeyTryer While I'm not persuaded by it, it's a closer analogy 👍

[Matt Jensen @MattJensenSea]

@Scobleizer @fchollet @KeyTryer Which of your examples is most similar to a globally-connected system trained with >10^24 MFLOPS of data, resulting in emergent powers of unknown scope, followed by end of 2023 w/ GPT5? Can you offer a tighter analogy? This is Apple IIs and oranges.

[Matt Jensen @MattJensenSea]

@RadioFreeTom Good times!

[Matt Jensen @MattJensenSea]

@bigjondaniel @KevinMKruse I gnew what you did there

[Matt Jensen @MattJensenSea]

@emollick Bostrom's typology seems to miss the hazard of legal liability. An officer/employee might prefer not to learn of facts that obligate them to act. It falls closest to "embarrassment" or "disappointment", but possibility of charges is more concrete, arguably stronger, than those.