BIG AI News : Open Source CRUSHES Everything, GPT-5 Paramters Leaked, AGI Could BeDecades Away?

Prepare for AGI with me – https://www.skool.com/postagiprepardness
π€ Follow Me on Twitter https://twitter.com/TheAiGrid
π Checkout My website – https://theaigrid.com/
00:00 – Introduction to Reflection 70B open source model
00:51 – Discussion of model performance and benchmarks
02:09 – Explanation of reflection tuning technique
04:34 – Implications of Reflection 70B’s performance
05:18 – OpenAI considering higher-priced subscriptions
07:56 – Discussion on the value of AI models for businesses
10:22 – Interview clip about AI cost reduction and improvement
11:32 – Xdai team’s Colossus AI training system
13:01 – Andrew Ng’s comments on AGI timeline
15:23 – Debate on AGI definition
16:33 – Ilya Sutskever’s new AI startup focused on superintelligence
18:18 – Speculation about GPT-5 model size
19:39 – One X Robotics CEO on scaling robot production
22:20 – Introduction to Google’s AlphaProteome
24:10 – SEAL leaderboard results for AI models
Links From Todays Video:
https://x.com/elonmusk/status/1830650370336473253
https://x.com/tsarnick/status/1830175922411966755
https://x.com/TheHumanoidHub/status/1830331686653149628
https://x.com/tsarnick/status/1830045611036721254
https://www.reuters.com/technology/artificial-intelligence/amazon-turns-anthropics-claude-alexa-ai-revamp-2024-08-30/
https://x.com/GoogleDeepMind/status/1831710991475777823
https://x.com/mattshumer_/status/1831767014341538166
https://www.theinformation.com/articles/openai-considers-higher-priced-subscriptions-to-its-chatbot-ai-preview-of-the-informations-ai-summit?utm_campaign=Editorial&utm_content=Newsletter%2CAI+Agenda&utm_medium=organic_social&utm_source=twitter&rc=0g0zvw
https://www.reuters.com/technology/artificial-intelligence/amazon-turns-anthropics-claude-alexa-ai-revamp-2024-08-30/
Welcome to my channel where i bring you the latest breakthroughs in AI. From deep learning to robotics, i cover it all. My videos offer valuable insights and perspectives that will expand your knowledge and understanding of this rapidly evolving field. Be sure to subscribe and stay updated on my latest videos.
Was there anything i missed?
(For Business Enquiries) contact@theaigrid.com
#LLM #Largelanguagemodel #chatgpt
#AI
#ArtificialIntelligence
#MachineLearning
#DeepLearning
#NeuralNetworks
#Robotics
#DataScience
@ariisaac5111
October 12, 2024 at 5:29 am
@AIgrid, I love your daily AI news show, but I had to stop following you because it's too much hassle because YouTube has disabled the playoff feature for your videos and it's too much hassle to add your videos to the queue. Can you create a playlist on your homepage that plays all your shows from newest to oldest, so that I can easily keep up with you just by pressing play off I usually do it every few days or once a week. thx.
@jeiel2027
October 12, 2024 at 5:29 am
π― Key points for quick navigation:
00:00:00 π€ Modelo de CΓ³digo Aberto ReflexΓ£o 70b
– O ReflexΓ£o 70b Γ© um modelo de cΓ³digo aberto com 70 bilhΓ΅es de parΓ’metros que se destaca entre os modelos atuais, superando outros modelos de estado da arte, mesmo os maiores como o GPT-4.
– Utiliza uma tΓ©cnica chamada "reflection tuning" para auto-corrigir erros e melhorar respostas.
– Apresenta uma melhoria significativa nas habilidades de raciocΓnio e precisΓ£o em benchmarks importantes.
– O modelo estΓ‘ atraindo muito trΓ‘fego, o que pode limitar seu uso inicial.
06:07:00 πΈ PossΓvel Aumento no PreΓ§o das Assinaturas do ChatGPT
– A OpenAI estΓ‘ considerando aumentar significativamente o preΓ§o das assinaturas para novos modelos de linguagem, podendo chegar a atΓ© $2,000 por mΓͺs.
– O custo mais alto pode refletir a alta capacidade e a melhoria nos modelos, especialmente para tarefas complexas em empresas.
– O preΓ§o elevado pode ser justificado pelo custo de treinamento e operaΓ§Γ£o dos novos modelos.
10:03:00 π AvanΓ§os em SuperinteligΓͺncia e a VisΓ£o do Futuro
– DiscussΓ΅es sobre superinteligΓͺncia e o avanΓ§o rΓ‘pido da tecnologia sΓ£o destacadas, incluindo investimentos significativos em novos sistemas de treinamento de IA.
– A startup SSI, fundada por Ilia Satova, estΓ‘ focada exclusivamente em superinteligΓͺncia, sugerindo que pode estar mais prΓ³xima do que se pensava.
– A capacidade de treinar modelos de IA rapidamente e o crescimento exponencial esperado sΓ£o destacados.
13:03:00 π§ Debate sobre a Linha do Tempo para AGI
– Andrew NG sugere que a AGI (InteligΓͺncia Geral Artificial) estΓ‘ a muitas dΓ©cadas de distΓ’ncia, e que as previsΓ΅es mais curtas sΓ£o baseadas em definiΓ§Γ΅es nΓ£o padronizadas de AGI.
– A AGI, conforme a definiΓ§Γ£o padrΓ£o, deveria ser capaz de realizar qualquer tarefa intelectual que um humano possa.
– As diferentes definiΓ§Γ΅es de AGI e suas implicaΓ§Γ΅es sΓ£o discutidas, com Γͺnfase em como isso afeta as previsΓ΅es sobre o futuro da tecnologia.
18:15:00 π Detalhes sobre o GPT-5
– O GPT-5 pode ter de 3 a 5 trilhΓ΅es de parΓ’metros, dobrando o tamanho do GPT-4, mas o foco deve estar nas capacidades de raciocΓnio e confiabilidade dos modelos, nΓ£o apenas no tamanho.
– ComparaΓ§Γ΅es e especulaΓ§Γ΅es sobre o tamanho e a capacidade do GPT-5 sΓ£o discutidas.
– A relevΓ’ncia do tamanho do modelo Γ© colocada em perspectiva em relaΓ§Γ£o Γ s suas capacidades prΓ‘ticas.
20:01:00 π€ AvanΓ§os na RobΓ³tica Humanoide
– A empresa 1x Robotics planeja escalar a produΓ§Γ£o de robΓ΄s humanoides significativamente, com metas de aumento exponencial nos prΓ³ximos anos.
– O desafio da fabricaΓ§Γ£o em larga escala e as estratΓ©gias para alcanΓ§ar essas metas sΓ£o abordados.
– A importΓ’ncia de ter uma abordagem cuidadosa e organizada para alcanΓ§ar grandes volumes de produΓ§Γ£o Γ© destacada.
21:55 𧬠AvanΓ§os no Design de ProteΓnas com IA
– Alpha Proto, um novo sistema de IA da Google DeepMind, gera proteΓnas inovadoras para pesquisa em biologia e saΓΊde.
– A IA Γ© treinada para criar proteΓnas que se ligam a molΓ©culas-alvo especΓficas no corpo.
– Melhora na eficΓ‘cia e velocidade na criaΓ§Γ£o de proteΓnas comparado aos mΓ©todos antigos.
23:12 π AvaliaΓ§Γ£o de Modelos e Rankings
– DiscussΓ£o sobre os rankings de modelos de IA, destacando o desempenho do mistro large 2 e o desempenho abaixo das expectativas de GPT-4 em algumas mΓ©tricas.
– ComparaΓ§Γ£o de modelos em diferentes Γ‘reas como codificaΓ§Γ£o e matemΓ‘tica.
– ObservaΓ§Γ£o de que GPT-4 foi superado por outros modelos em mΓ©tricas privadas e especulaΓ§Γ£o sobre quando a OpenAI pode recuperar sua posiΓ§Γ£o de lideranΓ§a.
Made with HARPA AI
@olternaut
October 12, 2024 at 5:29 am
AGI is not decades away. It's going to hit soon and like a freight truck!
@crumply
October 12, 2024 at 5:29 am
Reflection ended up being a total scam.
@jtabox
October 12, 2024 at 5:29 am
0:30 welp… this aged well i guess
@OmarMcIver
October 12, 2024 at 5:29 am
The start with Reflection 70b didn't age well.
@RunnerProductions
October 12, 2024 at 5:29 am
Definitely coming to this channel frequently to be stunned, surprised, and staggered
@regandediana
October 12, 2024 at 5:29 am
2K a month creates an extremely high barrier to entry for the average middle class (individual or family) to access the technology. It almost becomes exclusive to upper class/elite, increasing the gap between low/middle to upper/elites. terrible news.
@MatthewPendleton-kh3vj
October 12, 2024 at 5:29 am
More blatant class warfare.
@JoeyBlogs007
October 12, 2024 at 5:29 am
The thing is that the permutations are exponential with only a moderate increase in AI model size.
@BryanTyrone-l5c
October 12, 2024 at 5:29 am
Jones Brenda Johnson Elizabeth Martinez Sarah
@JoeyBlogs007
October 12, 2024 at 5:29 am
AGI is a bit like fully autonomous driving. They claim it's just around the corner, but in reality it's always a decade away. The closer you think you are, in reality the further away it actually is. π€£π€£π€£
@viralferrets1323
October 12, 2024 at 5:29 am
6:25 i donβt mind paying $2000/ month if this AI model would make me $5500/ month. π
@elia0162
October 12, 2024 at 5:29 am
so elon musk about agi was right?
@alexanderlinderson2655
October 12, 2024 at 5:29 am
Idk what David Matta is on about. it's a GENERAL intelligence. He's talking about creating "multiple AGIs" presumably that has specific skills. Well then it's not AGI, is it?
@yolxanderjacagonzalez5177
October 12, 2024 at 5:29 am
Cyberpunk future is on the way
@paulmclean876
October 12, 2024 at 5:29 am
…not very surprising really… inference times need to be applied to nake it better but all good… additionally, hiding the unnecessary working will further improve it… seeing the steps isn't useful most of the time…
@UnknownMoses
October 12, 2024 at 5:29 am
Agi is a bad thing donβt be a fool
@Adam-nw1vy
October 12, 2024 at 5:29 am
His name is Andrew pronounced (E)ng. The Ng part isn't an abbreviation.
@RanHab
October 12, 2024 at 5:29 am
guys i'm just starting out as an AI enthusiast,
would love your feedback as i make similar stuff!
@diaitigai9856
October 12, 2024 at 5:29 am
This video brings an interesting and more general approach to artificial intelligence; especially with the presentation of a new model, the Reflection 70B. This is quite amazing to see that open-source models are advancing and threatening the control of closed-source systems. The reflection tuning technique is the best of all the presented techniques as it provides information on how AI can update and improve its reasoning processes. Further, an analysis of the prospective pricing mechanisms of the OpenAI sheds light on the dynamics of the AC accessibility and value proposition. It is kind of thrilling to see such advancements happening now and it is even more fascinating to contemplate how these advances will define the future of artificial intelligence and the manner in which it will impact different industries. A great update with a lot of details, thanks for the info
@raul36
October 12, 2024 at 5:29 am
I find it funny to read comments from people who have no idea about artificial intelligence saying that Andrew Ng has no idea what he is talking about, when he is one of the greatest scientists in the field of artificial intelligence. I love watching how people talk nonsense when other people's lines of thought don't fit their parameters. In that case, everything others say is wrong.ππ
@TheAIExplorer-o5j
October 12, 2024 at 5:29 am
That transition at 13:06 was seamless! Your editing skills are seriously impressive.
@kelvinherbst
October 12, 2024 at 5:29 am
Basically, I'm truly truly listening to this at 2x speed.π
@damienjones9667
October 12, 2024 at 5:29 am
βAnd I think, and I think, and I think, and I think, and I think, and I think, and it think.β
Yeah well you donβt know. Delete your channel bro