Based on the 'behavior cloning' session, when we do RAG (Retrieval-Augmented Generation), is the training process teaching the model to output the correct answer for questions it doesn't have direct knowledge of, potentially leading to hallucinations? Or, when we perform RAG, do we typically provide more informative documents rather than just FAQ rows, giving the model much more context to understand the full scope of the knowledge, rather than simply guessing the type of question?
@irshviralvideo20 күн бұрын
slides link please
@kametor24 күн бұрын
Thank you sir for sharing this video, future is exciting 😃
@zuxiana29 күн бұрын
I don't think I have ever posted a YT comment in my life, but how this video has only 3 likes????? This is a amazing series, thank you for sharing this!
@nowithinkyouknowyourewrong8675Ай бұрын
This was more interesting than Dans talk but we still got Dan reading it
@hzheng9833Ай бұрын
公派留学生😂
@samwightАй бұрын
There's a weird personification of language models that nobody does for image generation models. It's weird to assume that these things "think", that they "do" things, when they're fundamentally no different than image generation models: they generate text that sometimes happens to be right. It's just that we've trained these models to *pretend* like they're intelligent.
@redoktopus3047Ай бұрын
he has that tweet up there at 3:06 from Sam Altman who says no data submitted to OpenAI is used to improve the models. That's just a flat out lie and Eric Wallance knows it. He has a paper "Poisoning Language Models During Instruction Tuning" where in the abstract he says "OpenAI leverages examples submitted in the browser playground".
@PaulGavrikovАй бұрын
What a great lecture! Love the real world poisoning examples
@420_gunnaАй бұрын
hendryck lamar
@420_gunnaАй бұрын
The bit at 32:45 about Szegedy being discouraged from publishing about adversarial examples by Ng is some interesting lore if true
@420_gunnaАй бұрын
sippin da henny
@simplicity84022 ай бұрын
Congratulations!
@senatorlainez2 ай бұрын
Really sad that people still share Charles Duell's apocryphal "quote" over 100 years later. In fact, he said this in 1902: "In my opinion, all previous advances in the various lines of invention will appear totally insignificant when compared with those which the present century will witness. I almost wish that I might live my life over again to see the wonders which are at the threshold."
@carvalhoribeiro3 ай бұрын
He explains very clearly, and the pauses allow me to organize the ideas into a flow, making it easier to follow the topic. Thanks for sharing this
@Thebetterlife_99913 ай бұрын
감사합니다. 교수님
@SUZAKU__0073 ай бұрын
Seeing this after gpt-4o 😊
@SM-qo9gr3 ай бұрын
my boy Jimmy is swollen af
@temle74893 ай бұрын
Got into EECS.
@gabrielmoreno32743 ай бұрын
yooo good job! same
@DJTrancenergy3 ай бұрын
It's amazing that a professor of a top university, no less, thinks this will be a game changer. It'll hardly speed up the easy things in a SoC. Seems a rather naive idea to me.
@chefatchangs48374 ай бұрын
Jim Keller if fucking jacked hahaha
@mondomonkey85814 ай бұрын
𝙼𝚘𝚜𝚝 𝚎𝚡𝚌𝚎𝚕𝚕𝚎𝚗𝚝.
@isabellaborkovic24484 ай бұрын
this dhruv fellow has charmed me
@azamatbezhan16534 ай бұрын
Stacked LEGO-PoL CPU Voltage Regulator . What do you think about it.
@sethcoast4 ай бұрын
I am very excited by this line of research
@waterblah61674 ай бұрын
1:00 dawg 💀
@18000rpm5 ай бұрын
Best subtitling I've seen in any video.
@marccawood5 ай бұрын
The supposed benefits are extremely naive. Supposing you got all state for a huge distributed OS in one database… that does not mean any application can read any data - which is what you require for many of the supposed benefits including monitoring - this would be a security nightmare so you have to lock everything down with complex auth systems. Also the supposed OAOO benefits assume debit and credit DBs are of the same company and would therefore be moved to one VoltDB. They are not.
@metalim5 ай бұрын
video is interlaced? wtf? replace potato with a camera already
@İmmortallvikings6 ай бұрын
I am proud of with Dr.Shanaz Shabazov. Because she is the pride of Azerbaijan ❤
@grimsk7 ай бұрын
Request captions!
@grimsk7 ай бұрын
"이 이후로 apple 뿐 아니라 meta, ms, amazon 같은 비반도체 회사들도 저마다 ai칩을 만들기 시작"
@won20529jun7 ай бұрын
Amazing! I had no idea there was already so much progress in this area
@wege84097 ай бұрын
Great lecture, but there is some kind of a buzzing, is there any way to EQ the audio?
@utube38057 ай бұрын
Michael Stonebraker nice to see you again. Hope to hear more.
@tejas818 ай бұрын
What a wonderful story teller.
@royalfalcon20218 ай бұрын
There is some high frequency sound is coming from the video I don't know if anyone able to listen it
@royalfalcon20218 ай бұрын
There is some high frequency sound is coming from the video I don't know if anyone able to listen it
@adolephasmith61168 ай бұрын
IT IS A BIG NEWS, thanks for sharing.
@koushikdey86098 ай бұрын
Very nice talk....
@redSTEM8 ай бұрын
Claire Tomlin is amazing
@gagegr8 ай бұрын
Thank you for sharing this with the world!
@idfubar9 ай бұрын
@32:08 "'Turning Brain Waves Into Words'"
@AlgoNudger9 ай бұрын
*What's the clear definition of (human) intelligence?* 🤔
@user-ee8te3dh2b10 ай бұрын
Fascinating
@AlgoNudger10 ай бұрын
❤
@Shiffo10 ай бұрын
57:05 I think Lex Fridman broke the record of a half an hour conversation with codec avatars.
@ctankep11 ай бұрын
" Oh brother where art thou "
@ceceropanini664411 ай бұрын
So beautiful ❤️❤️👌👌 However, people like Rana Kar, Sudipto Pal, Aniruddha Dasgupta are criticizing these beautiful ventures saying that industries should not embrace trade quantum computing, AI investment. They advocate for old system of work.