Software developer passionate about NLU. Canadian. Enjoys following SpaceX, Tesla, etc. Appalled by disparity between rich and poor on planet earth. INFJ.
@benwiley4000
@AndrewHartAR
You can use the "do we really need" argument for any one of thousands of inventions that improve the efficiency of a task. All other things being equal, efficiency has value. If your point is to wonder if efficiency can be improved by AR for things like this, definitely.
@WholeMarsBlog
@elonmusk
As a software dev, it would be pretty surreal to be at the office, huddled around the richest dude in the world (who is a coder at heart) brainstorming next steps.
@elonmusk
What I love the most is how quickly this video was released -- and that it was released at all -- really takes us along on the adventure. Thank you!
@skirano
We joke, but the thought roughly crossed my mind today: When will LLMs start becoming part of the feedback cycle that helps propel ML forward even faster.
@EndWokeness
I don't understand this take. Over time, economies transform, and areas that were once economically viable may become unviable. Pouring money into an unviable area to try to stop the inevitable isn't necessarily the right thing to do.
@soumithchintala
Fascinating! This would actually be another reason perhaps for them not to be too transparent about their architecture, if it gives off the "we're a bit low on ideas" vibe.
@elonmusk
@cabbey
@JeffMcAdams
Idea: Create a Tesla website where people can nominate voice commands or missing wordings and owners can upvote/downvote.
@angryUTdude
@SciGuySpace
The crazy thing is, Jeff has way more than enough money to do that, and this is his life dream (ish). Seems weird that they priced themselves out of the market.
@SERobinsonJr
It's less hard for me to believe that they could learn, accidentally, that swimming quickly under an ice sheet causes it to crack, but the fact that they could obviously plan, communicate, and coordinate with each other well enough pull it off is... 🤯
@SciGuySpace
Also: We've wondered what a Mars landing will look like, and whether it would be SpaceX, or SpaceX+NASA. With this news it's looking like NASA would quite likely work with SpaceX to land on Mars using Starship.
@sama
It did shift for me the likelihood that we'll see something equivalent to AI passing a Turing test in a meaningful way in the next 10 years. That might be an overreaction, but it's starting to feel like the writing is on the wall.
@meghamama
Every person gets 24 hours a day. How much economic value they produce during those 24 hours varies greatly. Ironically what you are observing implies the opposite of what you're concluding.
@lacker
When I read the line "An interesting corollary is that GPT-3 often finds it easier to write code to solve a programming problem, than to solve the problem on one example input" my brain said "oh crap, I wonder if that's what the human brain does, and has an executor".
@amuseddaman
@Cinnabar233
I heard that giving the JEE test to GPT was flawed, since it did not include the diagrams from the questions. Do you know if that's really the case? And did any of these IIT questions involve diagrams?
@roydanroy
There's a certain irony that ML has become quite "hot" in the last decade, and yet I feel like ML is sort of the "long game", requiring an attention that spans decades. A danger of the hype is that it makes it look like a gold rush, switching our minds out of long-game mode.
@tsarnick
Additionally:
- Might be part of a search algorithm for goodness: Once you find sufficient starting conditions for a universe, let it play out, and select the parts that are deemed good.
- For example, individual humans might be those good things being discovered / selected.
@elonmusk
I just realized that the RCS thruster on the falling side probably thrust its little heart out as the stage tipped over, possibly keeping velocity of impact low enough to keep it in one piece.
Aside from experimenting with GPT-3 to see what it can and cannot do, the lasting thing that I enjoy the most is asking it questions. I find its answers thought provoking, almost wise.
@AmandaAskell
Timnit has made this claim in the past about Will and EA, but I've noticed it's not uncommon for her to make a lot of noise and slander people with things that are both factually incorrect but very serious accusations. I find her to be quite irresponsible.
@RollingStone
One of the things I like most about Elon is that he highly values the truth. As he says, don't read between his lines. Such a big asset and differentiator in a world of spin. Second, I admire his openness about his emotional pain. Hard not to like the guy.
@TSLAFanMtl
Dropping prices may make investors wince, but IMO that's ultimately what we want as citizens of earth: For EV prices to come down, and with it, significantly more people being able to afford them, which keeps the Wright's Law motor spinning.
@elonmusk
@adamchavez
Thank you Elon for a formal apology. For people like me that deeply respect your efforts of engineering and heart, reading your comments to Mr. Unsworth was unsettling and saddening. Your admission of missing the mark is helpful.
@getjonwithit
I've read a couple of posts from you in the last week that refer to others using very negative language. If it makes any difference, I find tweets like that quite off putting, possibly because to me it gives the impression of arrogance.
@arstechnica
@SciGuySpace
All of Elon's talk about the craziness of throwing away a 747 after each flight is turning into very real changes across the space business. Success.
Given that
#gpt3
obviously has a very detailed model of language and grammar, I was curious to see if it could both correct grammatical errors and explain the corrections. The answer is "yes", although it took more retries than I thought it would for explanations:
@FeepingCreature
Well, what it means precisely to "understand" is a bit uncertain. People might be using the same word in different ways. I agree that GPT-3 doesn't measure up anywhere close to human understanding, but I do think a Venn diagram with "understanding" and GPT-3 should show overlap.
@elonmusk
@JohnnaSabri
@dtemkin
Reading the needlessly discouraging and borderline malevolent things people have written about your efforts on the sub project make me sad. I was very delighted and inspired by you and your team.
@matthen2
I've long imagined that something like this happens when we try to connect two things in the brain, although perhaps more precisely like Dijkstra's algorithm -- sending out two frontiers until they meet, triggering a little lightning strike.
@Rainmaker1973
There seems to be some profound symbolism here -- within human societies that lose their "true north", where each of us is "following the crowd", might we succumb to an analogous fate?
@JefRexx
@NASASpaceflight
Felt a bit choked up as well. I'm so used to NASA making weird choices that are politically motivated but technically and financially bizarre. Seems so hopeful that things finally seem to be coming into alignment in terms of NASA making smart choices.
@peterrhague
One thing I heard for the first time today: Most Starships won't return to earth, they will be used for their materials on Mars. That really changes the math on net cost per trip to Mars, because each trip doesn't allow you to re-use the ship.
@Cmdr_Hadfield
Once upon a time there was a guy traveling back from his first real vacation abroad (Europe) so that he could preside over this hotel's opening. Unfortunately his ride back to North America was the Titanic.
@kimpaquette
@Kristennetten
I've found this with some people in my family as well. Because I find Tesla interesting and inspiring, it's natural for me to share that interest, but sometimes it feels like people are more than disinterested, they seem to "actively" dislike the topic. Sad.
@kimpaquette
@Kristennetten
Could watch these all day. 2020 has been a rough year for the world, but this is a part of 2020 that has been really interesting to experience, if vicariously.
@lexfridman
@elonmusk
As some have tweeted: "I'm never deleting this app" ... pretty astonishing to watch this kind of exchange happen in real time, out in the open. It reminds me of the Truman Show somehow.
@DevinOlsenn
Great video! I feel like 12.5.x is significant for two reasons:
1. Seems to increase confidence significantly that we're on the path to true FSD, robotaxi.
2. But also, I think FSD just passed the threshold to something that millions of people will want to pay money for.
@FitFounder
Repeatedly do the following:
(1) Take an important area of life: Diet, exercise, relationships, mental health, work, etc.
(2) Consider a simple pattern that you could experiment with to improve.
(3) Try it for three weeks, and evaluate.
@stanislavfort
@AndrewSteinwold
I think the idea is that, if params are that high, it is a sparse model, but in that case, saying it's hundreds of times more powerful is incorrect. (if I'm understanding correctly)
@sai_prasanna
I strongly agree. The sentiment that GPT-3 was a terrible misuse of natural resources was one of the weakest points made against it, IMO. Probably even more so if that point is made against AlphaFold.
@_akhaliq
This is super cool, but if I'm understanding correctly, this is BLITZ, where there isn't much time at all to think, so you need to use instincts. Thus it makes sense that search isn't as important.
@togelius
@GaryMarcus
Although GPT-3 is perhaps best characterized as an association engine (ish), the emergence of its few-shot adaptation ability was a huge surprise for me and makes me curious whether there will be any more surprises as it is scaled up further.