Пікірлер

• 
  @George___340k11 күн бұрын

  Catch an exclusive interview with Binance’s CEO to get the scoop on future developments

• 
  @yimingyang425412 күн бұрын

  Due to some AV issues, I could not hear the questions clearly on the stage. So, let me clarify some of the answers retrospectively. Question 1. Why did the easy-to-heard voting strategies perform worse than the SFT baseline with greedy decoding when the candidate-pool size is less than 10? Answer: The voting strategies had a non-deterministic process, with the temperature set to 0.7 (as shown in the slide) while the baseline has the temperature set to 0 (deterministic). This means that the baseline always picked its top candidate per problem instance, but the voting strategies may miss it when the pool size is rather small. Question 2. How does the proposed method differ from GAN? Answer: The discriminator in GAN is trained to label each instance as a natural-language output (yes) or un-natural (no), while the evaluator in easy-to-hard generalization is trained to tell whether a math solution is correct (yes) or wrong (no) for a given math problem. Even if an answer is perfect in English, it still can be wrong mathematically. Question 3. Can we use this idea to improve LLM pre-training? Answer: Maybe not. When we have enough (unlabeled) data for pre-training of an LLM, we may not be benefited much from the easy-to-hard generalization. On the other hand, if we do not have enough data for pretraining an LLM, we may also not be able to train the evaluator well. One may argue that what if we can train the evaluator on rare patterns that the current LLM cannot handle well? Perhaps yes, but this is a big “if”. That is, the challenge is then shifted to how to obtain the annotated data on the rare patterns. I hope the above answers help.

• 
  @joeokeefe277112 күн бұрын

  How did I get here

• 
  @AlgoNudger14 күн бұрын

  Thanks.

• 
  @Roibarkan15 күн бұрын

  8:59 Stream starts

• 
  @suyashsingh108520 күн бұрын

  Worst workflow. Very frustating

• 
  @j.r.366421 күн бұрын

  Congratulations! A well-deserved prize.

• 
  @jesalgandhi25 күн бұрын

  legend

• 
  @user-ub5ok1dq9p28 күн бұрын

  real man !

• 
  @djethereal9929 күн бұрын

  great paper!

• 
  @ats89117Ай бұрын

  This new workflow template is a mess...

• 
  @maximilian19931Ай бұрын

  youtube limits download rate server side to about 1.2 times the bitrate of the video by setting a cwin flag on the TCP socket!

• 
  @user-ji6ze7gn1qАй бұрын

  Asesome Prezzo❤, Is Moka an open source tool?

• 
  @PraetorClaudiusАй бұрын

  Unfortunately, this is a glimpse of the horror of The age.

• 
  @mostafatouny8411Ай бұрын

  Thanks for publicly sharing the talk; Avi is inspiring for all generations.

• 
  @MaximSchuermansАй бұрын

  Introduction at 6:40 and talk starts at 10:45.

• 
  @faisalsheikh7846Ай бұрын

  Amazing❤

• 
  @AlgoNudgerАй бұрын

  Thanks. 😊

• 
  @java_Marcelo-xx5nwАй бұрын

  Thank you for share!

• 
  @AaronNGrayАй бұрын

  Yes I learn lots from the original Minix book all those years ago. Thank you Andrew !

• 
  @RoibarkanАй бұрын

  28:51 Kostya’s talk: kzread.info/dash/bejne/o4ifu6Z6YKjUZrg.html

• 
  @ledasdaughter_Ай бұрын

  This is fascinating and I learned so much. I was only marginally aware of those branches of research and the developments, but it's so good to be able to listen to these wonderful people sharing their research and the important work they do. Absolutely fantastic and so very important.

• 
  @ASmile-xk1efАй бұрын

  Interesting panel. I love that research like this is being done. So important.

• 
  @bilimtv791Ай бұрын

  No to lgbt

• 
  @roamingeek1129Ай бұрын

  Funny that Tulsee Doshi was on here talking about Google and fairness, only to have Gemini shown to have the most egregious bias ever seen in Artificial Intelligence.

• 
  @VolumetricTerrain-hz7ciАй бұрын

  There are unknown way to visualize subspace, or vector spaces. You can stretching the width of the x axis, for example, in the right line of a 3d stereo image, and also get depth, as shown below. L R |____| |______| TIP: To get the 3d depth, close one eye and focus on either left or right line, and then open it. This because the z axis uses x to get depth. Which means that you can get double depth to the image.... 4d depth??? :O It's even possible to add 4d depth to a isometric cube, by moving the skew square side along the x axis of the right cube, individually, and leave the left cube unchanged. Would be awesome, if someone make a game (even KZread video) that way! Nice video! :-)

• 
  @FREELEARNINGАй бұрын

  Very nice and profound Talk. Thanks Hannes and Thanks for ACM for sharing this publically on KZread.

• 
  @AlgoNudgerАй бұрын

  Keswick.

• 
  @biplavoraon9702Ай бұрын

  Wow! This is such a good introduction to the field. Thank you for this❤

• 
  @daxteam5719Ай бұрын

  Creo que se paso de ver... con lo de sasuke

• 
  @FernandoYtusaca2 ай бұрын

  🎯 Key points for quick navigation: Understanding anti-patterns in student code can indicate knowledge gaps and lack of understanding of correct patterns. Code writing alone may not provide a complete picture of a student's knowledge, as students can generate code without understanding underlying principles. Students who use anti-patterns in code writing may not necessarily have conceptual gaps, but may lack understanding of specific patterns. Code writing is weakly correlated with editing success, suggesting that additional tasks besides code writing may be necessary to assess student knowledge accurately. Engaging students emotionally with real-world examples, such as accidents caused by suboptimal software testing, can enhance learning and motivation. Learning from accidents in software engineering can help students understand the importance of safety and domain knowledge in software development. Encouraging witnesses of accidents to share their experiences with software engineering students can lead to better education and training of future software engineers. Students in the University course on mlops are taught how to adopt mlops practices and gain practical experience with mlops tools through Project based learning. The course is organized into six milestones focusing on different aspects of mlops, such as reproducibility, quality assurance, API development, containerization, and monitoring. The pilot run of the course received positive feedback, with all students successfully deploying their machine learning components to the cloud. The software Resurrection exercise helps learners understand software maintenance by building and testing historical software releases, fixing issues, and reflecting on the experience. The exercise allows learners to experience the challenges and joys of software maintenance, contributing to readability, intelligibility, and maintainability of program code and documentation. Teaching programming patterns can be done at the later part of an undergraduate course Students should have a background in compiling software and programming in a language like C Learning to write programs that are easy to understand and maintain is valuable for both students and professionals Made with HARPA AI

• 
  @marlhex62802 ай бұрын

  I want a picture with bob! and an Ether!

• 
  @Thebetterlife_99912 ай бұрын

  대단하신분🎉

• 
  @inkbunnybunny2 ай бұрын

  I have a question would it be possible to have more than just User and Kernel space?

• 
  @didyoustealmyfood87292 ай бұрын

  Look at my man yann. so inspirational...

• 
  @maxmustermann55902 ай бұрын

  Great work, wish the talk were longer

• 
  @gonzales4202 ай бұрын

  interesting.

• 
  @user-tc7gm7cl8r2 ай бұрын

  Thank you for sharing. I can not wait to check this paper.

• 
  @vaibhavrathore82512 ай бұрын

  This is the most comprehensive tutorial till date.Great work .

• 
  @hermis20082 ай бұрын

  I would love to know why there is poorer quality on Linux than in Windows. In Ubuntu I get a 1.117Gbps while on Windows I get 6.84Gbps.

• 
  @JaneSoap-oc8vu2 ай бұрын

  You are dilusional!! You are hurting people. Your children and theirs will pay for what you are doing for money And research. I pray for your souls

• 
  @RozarSmacco2 ай бұрын

  Will the public ever be allowed to acknowledge or accept sociobiology?

• 
  @panshidong84672 ай бұрын

  The description seems not match the video content

• 
  @mahkhi71542 ай бұрын

  I wrote Tetris in 1 Month, that's because I Wrote it in C. Those Borgs Spent 2 Years Writing Tetris, because they had to Do it in Assembly. EVEN THEN: it would Take Them Borgs, Half a Dozen Borgs, Over 6 Months in C. The 286 executes C, without Having the Compiler, to Break the Code All the Way Down to X86 ASSEMBLY. The 386 executes C++, without Having the Compiler, to Break the Code All the Way Down to X86 ASSEMBLY. They Actually Reverse Engineered the 286, Which Ran Windows. Partially Reverse Engineered the 386. Made their Own Windows NT Like, Operating System.

• 
  @mahkhi71542 ай бұрын

  How DO I Know: The Minions (Redhat/IBM/Oracle) Gnome/GTK is written in C. Your God: Nothing Better Than C. They (Minions and Borg), Partially Reverse Engineered the C++ 386. The Borgs then Made their Own C++ KDE/Qt. The Minions and Your God, don't have Pure Gold like us Cats. They Cant Make Fast Processors. They Have to Rely on Others Processors. USE ARM CPU. Don't Accelerate Software in Hardware.

• 
  @lakpatamang28662 ай бұрын

  Great tutorial. Very informative.

• 
  @sabinalama62113 ай бұрын

  Is this man still alive now coz I just heard yesterday that he was no more.. and now I'm still searching for his news..

• 
  @chanukyasai2729Ай бұрын

  is it ture??

• 
  @sabinalama6211Ай бұрын

  @@chanukyasai2729 i think so

• 
  @ShortLessonsDaily3 ай бұрын

  Thank you for this video

• 
  @djdjmbenga3 ай бұрын

  how did this lady get this job lol

• 
  @fubarace10273 ай бұрын

  Should be outlawed in the US as discriminatory? Especially in state funded schools.

• 
  @colevq3 ай бұрын

  Perfect video for my project I got assigned!! Thank you