What are Large Language Models (LLMs)?

Поділитися
Вставка
  • Опубліковано 12 січ 2025

КОМЕНТАРІ •

  • @GoogleDevelopers
    @GoogleDevelopers  Рік тому +13

    Subscribe and stay tuned for the next episode! → goo.gle/developers

    • @AMFOREVIEW
      @AMFOREVIEW Рік тому

      ❤❤ Dear Developers,
      It's Okay ❤❤

    • @reasonerenlightened2456
      @reasonerenlightened2456 Рік тому

      ​@@AMFOREVIEW Deep Learning and GPT-4 are fake intelligence. For example, It struggles with fingers, and with drinking beer. LLM are a dead-end for AGI because they do not [understand]! the implications of their outputs! Also, GPT-4 is designed by the Wealthy to serve their needs!
      ------
      Douglas Lenat wrestled with creating the true AI . AI lost a giant.
      He was not correct about some AI-things but he opened my mind about great many more things about AI.
      RIP Douglas Lenat, 1950-2023

  • @CaitlinGary
    @CaitlinGary Рік тому +9

    Thanks for diving in! What a great explanation.

  • @spinnerlive
    @spinnerlive Рік тому +9

    The first AI transformer model, known as "Transformer," was introduced by researchers at Google in the paper titled "Attention is All You Need," published by Vaswani et al. in 2017.

  • @finnchristianarctander4414
    @finnchristianarctander4414 Рік тому +2

    🎯 Key Takeaways for quick navigation:
    00:00 🤖 Introduction to Large Language Models (LLMs)
    01:26 🧠 How LLMs Work
    03:19 📝 Prompt Design for LLMs
    04:43 🤯 Challenges in Writing Model Prompts
    05:13 🌟 Conclusion and Applications
    Made with HARPA AI

  • @e-genieclimatique
    @e-genieclimatique Рік тому +42

    In brief (GPT4):
    In this video, the speakers discuss the power of Large Language Models (LLMs) in understanding and generating human language. LLMs are based on the transformer architecture invented by Google and are trained on massive text datasets. Their ability to be used for a variety of tasks, such as chat, copywriting, translation, summarization, and code generation, makes them incredibly powerful and efficient.
    LLMs can be utilized without being a machine learning expert, as they function like sophisticated autocomplete systems. Users can input text and receive output text based on the patterns and language learned by the LLM. The input text is called a prompt, and prompt design is crucial for getting the desired output from the LLM.
    There are two main approaches for prompt design: zero-shot learning, which involves using a single command, and few-shot learning, which includes providing examples. However, there's no optimal way to write model prompts since the results are highly dependent on the underlying model, and small changes in wording or word order can have a significant impact.
    Users can try out LLMs like Google Bard and experiment with different prompt structures and formats to find what works best for their specific use case.

    • @kazirabbany4420
      @kazirabbany4420 Рік тому +1

      You are not using BARD? :P

    • @reasonerenlightened2456
      @reasonerenlightened2456 Рік тому

      GPT-4 is fake intelligence. For example, It struggles with fingers, and with drinking beer. LLM are a dead-end for AGI because they do not [understand]! the implications of their outputs!
      Also, GPT-4 is designed by the Wealthy to serve their needs!

  • @StephSancia
    @StephSancia 8 місяців тому +5

    I found more understanding with this video vs one 12 times longer due to the extremely positively beneficial way this was presented in an extremely easy to learn visual manner so thank you very much indeed ❤️

  • @Emezdee
    @Emezdee 22 дні тому

    What a wonderful video. Simple and concise.
    Paying so much attention that I saw the "Email Foramtting".

  • @oryanol
    @oryanol Рік тому +6

    I use bard everyday and it helped my workflow increase significantly

  • @Maksimgordeev-bj4nt
    @Maksimgordeev-bj4nt 8 місяців тому

    Your insights are like gems that light up every conversation.

  • @lgmuk
    @lgmuk Рік тому +5

    Very good video! These llms are much clearer now!

  • @luis96xd
    @luis96xd Рік тому +10

    Wow, amazing video, everything was well explained! I really learn a lot from your videos, thank you so much 🤗🤗

  • @charlesopuoro5295
    @charlesopuoro5295 Місяць тому

    Thank for this Clear Explanation

  • @Eltopshottah
    @Eltopshottah Рік тому +11

    Can't wait to see what Google has up their sleeves with A.I

  • @cassianocominetti7784
    @cassianocominetti7784 Рік тому

    Amazing video gilrs! Thank you!

  • @VaibhavPatil-rx7pc
    @VaibhavPatil-rx7pc Рік тому

    Excellent detailed information, Highly recommend everyone

  • @Netryon
    @Netryon Місяць тому

    It begins with a possibility of SparQL data modeling and now as some Grant Maclaren breakthrough is getting its traction at conferences and eventbrite/meetup meetups. They could have done that for a while, but you yourself were busy RTU VEF. 0:50 Maybe Dovilė Šakalienė would accept this drawing. 2:09 something you missing here is not your own old video you wish to crop and you wished everybody to do it. 2:53 your exams are just as this, won't pass if you don't know how your own program does the verification. Trying anything as this would make you build your servers and server rooms. You may begin with one rack and just build that system to the size of the user computer - waterfall is not what you want, but it's your best option. 3:15 why chesapeak? Do you see the red nato woman in the crowd 65? Something as this 4:09 too hard to define.

  • @LearnMalayalam65
    @LearnMalayalam65 11 місяців тому

    Thank you! this has helped me learn more about this

    • @GoogleDevelopers
      @GoogleDevelopers  11 місяців тому

      It's our pleasure! Glad you enjoyed the video 😎

  • @malakbelabdi5066
    @malakbelabdi5066 Рік тому

    Very simple and rich video
    Thank you!

  • @MrJekyllDrHyde1
    @MrJekyllDrHyde1 Рік тому +1

    When i finally I get GenAIs to respond with what i wanted (often this will take multiple prompts, corrections), i ask it a simple question:
    > "Please give the prompt which would have generated the last output"

  • @RaviKrishnaSrivastava
    @RaviKrishnaSrivastava Рік тому

    Very good video about LLM and prompt engineering

  • @confusion-capture
    @confusion-capture Рік тому

    Thanks for this overview!

  • @hesamarki7256
    @hesamarki7256 Рік тому

    Thanks for your explanations.

  • @marcellemagnima1047
    @marcellemagnima1047 11 місяців тому

    So clearly,thank you very much

  • @leletejasr
    @leletejasr 23 дні тому

    Hello Google,
    I am new to English, but want to improve in analytical writing for the GRE, how can I learn like LLM?😅

  • @martinseal1987
    @martinseal1987 Рік тому +24

    Like Google bard ...................

  • @vgreddysaragada
    @vgreddysaragada Рік тому

    well explained..Thank you so much

  • @orobosatheeagle
    @orobosatheeagle 4 місяці тому

    Great explanation

  • @singhRajshree
    @singhRajshree 4 дні тому

    How does LLM perform on non structured languages. Like hinglish languages?

  • @AMFOREVIEW
    @AMFOREVIEW Рік тому

    I'm really with a question.
    So, waiting for the next video to learn, what do you think isn't it a hard job?

  • @ncroc
    @ncroc Рік тому +9

    I still don't understand how do you go from a language model which predicts the next word to a question answering system.

    • @MattiaCeccopieri
      @MattiaCeccopieri Рік тому +2

      Same

    • @augusto256
      @augusto256 Рік тому

      Because it's impossible to predict undefined input.

    • @zvxcvxcz
      @zvxcvxcz Рік тому +5

      They don't per se. It just happens to be that in common text an answer frequently follows a question, so the next predicted words happen to be those starting the answer. They never actually leave next word prediction.

    • @ncroc
      @ncroc Рік тому +2

      @@zvxcvxcz I see but those systems seem to form an answer using the related documents. Let's say neural nets somehow determine which paragraphs are similar to the question asked. Do neural nets also form the answer text or there is some other algorithm that forms the answer text?

  • @christianfaust5141
    @christianfaust5141 Рік тому +1

    Excellent intuitive video about prompt engineering for starters, thank you.

  • @monkeydluffy2063
    @monkeydluffy2063 Рік тому +2

    "and definitely let us know what you are building"
    being google, I'm pretty sure you already know that

  • @artus198
    @artus198 Рік тому +1

    Pretty sophisticated stuff is going on with AI. How is PALM different to BARD ?

  • @danqiu6413
    @danqiu6413 Рік тому

    Watson developed by IBM more than 10 years was very good in language understanding. What model was behind Watson?

    • @trendthatneverend4812
      @trendthatneverend4812 11 місяців тому

      According to the knowledge available on public internet, watson is not built on a particular ML mdel. A bunch of algorithm and technologies is used to implement it. For more reference you can visit given link.

  • @pleban
    @pleban Рік тому +1

    Bard isn’t currently supported in your country. Stay tuned! - that's all about G approach

  • @jeanpeuplu3862
    @jeanpeuplu3862 Рік тому

    Is LLM part of NLP? Is an LLM always an NLP model? Or can an LLM be another kind of model?
    Also, BERT is Transformer but not an LLM, right? Transformer can be LLM or not, right?

  • @amparoconsuelo9451
    @amparoconsuelo9451 Рік тому

    Can a subsequent SFT and RTHF with different, additional or lesser contents change the character, improve, or degrade a GPT model?

  • @MarcoPreialmini
    @MarcoPreialmini 6 місяців тому

    Very good video!

  • @JayMaverick
    @JayMaverick Рік тому

    Nailed the "silence, brand" vibe.

  • @sirbughunter
    @sirbughunter Рік тому +2

    Great video. You guys are very sympathetic and explained the concept of an LLM in a simple but understandable way!

  • @DiyTechHubOfficial
    @DiyTechHubOfficial Рік тому +1

    Thanks for shring

  • @ravishmahajan9314
    @ravishmahajan9314 Рік тому

    France is to Paris , Japan is to .......????
    So you mean to say there is a sentence in word corpus that is very close to above sentence. So my question is i doubt that above sentence is there in word corpus then how will it predict Tokyo. Is there any other technique also used?

  • @ingenieroluisfer
    @ingenieroluisfer Рік тому +2

    🇨🇴👋🙋‍♂👍🤝 Saludos desde la ciudad de Bogotá D.C.

  • @erikm9768
    @erikm9768 Рік тому

    ChatGPT is the elephant in the room. I didnt even know you can talk to Google Bard at the moment

  • @fear_less_2020
    @fear_less_2020 Рік тому +1

    Really helpful. Thank you for explaining it so well!

  • @majorangle
    @majorangle День тому

    How did you know I was living under a rock?

  • @matt.stevick
    @matt.stevick 4 місяці тому

    The world officially changed forever, Pandora’s box unlocked 2024. I’m an early adopter and investor so I’m retired and watching this play out now 🍿

  • @USONOFAV
    @USONOFAV Рік тому +1

    Can you open source your models?

  • @PabloPazosGutierrez
    @PabloPazosGutierrez Рік тому +69

    Everyone heard about this because of ChatGPT which they try to not mention because it's a Microsoft thing.

    • @zvxcvxcz
      @zvxcvxcz Рік тому +14

      Microsoft has licensed use, but it is not a Microsoft thing, it is from OpenAI.

    • @Ani-rw4ln
      @Ani-rw4ln Рік тому +2

      Entrepreneurship is dirty bro you gotta do anything to just sustain. No other option

    • @learnerboy2247
      @learnerboy2247 Рік тому

      Google had created the first Ai chat bot 2 years earlier than open ai... But launch it because of its harmful effect on people

    • @alexleo4863
      @alexleo4863 Рік тому +1

      Google developed most of base tech that were later used by OpenAI, tech like from word embedding algorithms to Transformer architecture, and they gave them away for free, but Google are not arrogant enough to use the word "Open";😅

    • @ryan18462
      @ryan18462 4 місяці тому

      @@Ani-rw4lnya try to be more like Google

  • @jeffg4686
    @jeffg4686 10 місяців тому

    what is a mud kiln?

  • @charlene6306
    @charlene6306 Рік тому

    I still can’t access Bard. Something about permission setting blah blah

  • @deepfuchsia7204
    @deepfuchsia7204 Рік тому

    I like that the video starts with an insult.

  • @VIDEOSANDREEL
    @VIDEOSANDREEL Рік тому

    Thanks ❤

  • @MaitraRoy
    @MaitraRoy Рік тому

    The output of LLM is always probabilistic and never deterministic.

  • @ohaleceiffel
    @ohaleceiffel 9 місяців тому

    Looks like a dead heat in a zeppelin race 👀

  • @banzai316
    @banzai316 Рік тому +1

    Would be nice to have diffusers models that can run on mobile. Diffusers is so powerful

  • @zvxcvxcz
    @zvxcvxcz Рік тому +4

    Please stop saying they understand. They don't, and saying it over and over are giving normal people a very warped view of what these models are capable of. And yes, the bigger they get and the more info they're fed, the tougher it becomes to demonstrate that they don't understand because the outputs between the traines model and someone with actual understanding become closer together, but it really is actually the case that these models as they work today most certainly do not understand anything.

  • @CarlosPerez-hu7ly
    @CarlosPerez-hu7ly 7 місяців тому +1

    it seams i've been living under a rock jhajhajha ::/

  • @ken.matinale
    @ken.matinale Рік тому

    The words chat and conversation imply spoken communication. However, neither Google Bard nor OpenAI chat GPT are capable of speaking or understanding spoken words. It's strictly text, which gets tedious pretty quickly. These reviews never seem to reflect that. Plus, my view so far is that it's like the Wizard of Oz: just someone behind the curtain trying madly to keep up the illusion.

  • @tslee_
    @tslee_ Рік тому +1

    Waiting to see Bard Open-sourced😎

  • @a0100888
    @a0100888 Рік тому

    Chapter: fuchsia learning

  • @imranhossain3504
    @imranhossain3504 Рік тому

    I'm with you

  • @dronadulaeswar906
    @dronadulaeswar906 6 місяців тому

    gosh they are beautiful

  • @TebogoMotlhale
    @TebogoMotlhale 9 днів тому

    Not bad, but it is more like "How to/you use or interact with LLMs" and a little LESS about "how LLMs work".

  • @theratscratch
    @theratscratch Рік тому

    yeah guys like google bard

  • @pubservices2649
    @pubservices2649 Рік тому

    I need LLM

  • @aaronjiang8964
    @aaronjiang8964 Рік тому +3

    Mention ChatGPT without mentioning ChatGPT

  • @jerryyuan3958
    @jerryyuan3958 Рік тому

    nice explaination

  • @AlbanBytyqi
    @AlbanBytyqi Рік тому

    In my experience Bard has been totally unreliable.
    Ask bard how many about football results and it gets it wrong.

  • @deeplearningpartnership
    @deeplearningpartnership Рік тому

    Cool.

  • @sata1207
    @sata1207 7 місяців тому +1

    This video demonstrates what's wrong with google lately..

  • @amparoconsuelo9451
    @amparoconsuelo9451 Рік тому +1

    Where is the math? Where is the compilable code? How much is a language model? What would be the hardware requirements?

  • @jebuskmiest
    @jebuskmiest Рік тому

    So you guys are pushing prompt engineering as a real job?

    • @duppy404
      @duppy404 Рік тому

      There won't be dedicated "prompt engineers", instead prompt engineering is simply a skill that developers now need to have.

  • @sensorintelegence3480
    @sensorintelegence3480 Рік тому

    Advance AI

  • @jopadjr
    @jopadjr Рік тому

    2.2k+...Thanks

  • @chetanrawatji
    @chetanrawatji 10 місяців тому

    ❤❤

  • @qinlingzhou8815
    @qinlingzhou8815 Рік тому

    Wow. LLM is a black box you can not do it. We can train it for all tasks. What a waste! :(
    One can get a pre-trained free model to build own vector DB with own data and search them.

  • @DeloresCastillo-k1d
    @DeloresCastillo-k1d 4 місяці тому

    Martinez Kimberly Jackson Brenda Lewis Mark

  • @MohamedzakymohamedIlsisi-vx1wt

    I"m sempel

  • @ÖmerFaruk-t6c
    @ÖmerFaruk-t6c 11 місяців тому

    Deneme yorum

  • @MonteaNoLipton
    @MonteaNoLipton 4 місяці тому

    AI is a sophisticated...parrot!

  • @tomas-333
    @tomas-333 Місяць тому

    The perfect girl doesn't. exi...

  • @VikramBhaskaran
    @VikramBhaskaran Рік тому

    😢I 😢

  • @FazalMalik-yx7pz
    @FazalMalik-yx7pz 10 місяців тому

    Aweful audio quality