Generative AI Weekly Newsletter — Issue #2
Hey Everyone,
Welcome to the January 2024 edition of our Generative AI newsletter! As we step into a new year brimming with possibilities, we’re thrilled to bring you the latest and most exciting developments in the world of Generative AI.
Our mission remains steadfast: to keep you informed, inspired, and engaged with thought-provoking content, insightful analyses, and the most recent breakthroughs in Generative AI.
Recent AI News
- OpenAI launches the GPT store — OpenAI has publicly released their much-anticipated GPT store. The GPT store mirrors the style of an app store and has some cool applications already. To access the store, login to your ChatGPT account and click the GPT store.
Some cool ones we liked are the following:
Cartoonize yourself — Turn your selfie into a Pixar-style animation.
Consensus — An AI research assistant that can search a library of 200 million pages.
Books — Personalized book recommender.
Your thoughts matter to us! We’d love to hear about your favorites. Feel free to share them in the comment section below
2. Autogen studio 2.0 — Microsoft has released a new version of Autogen Studio, their multi-agent framework. This latest version, with its revamped user interface, stands as the first no-code solution that truly impresses, allowing anyone to effortlessly craft multiple agents, each with their own distinct persona, tools, and goals.
3. Google releases a stunning Medical Conversational Diagnostic Agent (AMIE) — AMIE is a cutting-edge medical diagnostic agent designed to have a conversation with a patient, ask relevant questions, and make a diagnosis of their condition.
AMIE is a LLM-based system trained using self-play. Self-play is a training method where an AI system essentially ‘practices’ by interacting with itself.
AMIE was tested against human doctors on real examples and “significantly” beat the doctors in accuracy of diagnosis (see chart below) while having more meaningful, empathetic conversations with the patients.
While AMIE is still in R&D, this is a promising step in making healthcare cheaper and more accessible for all.
4. Text to Video generation- 2024 marks a significant leap in AI advancements with the introduction of text-to-video generation, exemplified by ByteDance’s latest innovation, the MagicVideo-V2 model.
See an example below. At its core, MagicVideo-V2 employs a complex pipeline that melds text-to-image generation, video motion creation, and frame interpolation, producing results that are both impressive and seamless. We are keen to see how this field evolves this year.
5. OpenAI launches new models and lowers pricing — Among other OpenAI news, the company released 2 new embedding models — text_embedding-3-small
and text-embedding-3-large
.
They also released a new GPT4 model which is better at code generation and some new tools to manage and monitor API usage. The company mentioned they are planning to release a new version of their flagship gpt-3.5-turbo
model in the coming weeks while also slashing its price by 50%! Go OpenAI!
6. Meta releases Code Llama 70B model — Meta released a new Code LLama model that is 70B in size. This model is the biggest open source code generation model and can generate code in Python, Java, PHP and C++. Its code generation accuracy is comparable to GPT4 model.
Interesting stories of the week
- MidJourney’s Niji V6 anime generator — Niji is an anime-specific AI model with extensive knowledge of anime, styles, and aesthetics from MidJourney.
- Google’s text to Video model — Lumiere — See results from Lumiere and a brief explanation of how the model works.
- Youtube is getting some crazy new AI features — Youtube is adding new tools for video creators like Youtube Create and Dream Screen. Very cool!
- CrewAI in Action: Building your dream team — Crew AI is a recent open source solution to build your multi-agent team, giving them roles, memory, and customizing how they interact with each other.
- Run 70B LLM model on a 4GB GPU — The future of Generative AI is running more with less.
- Google’s Geometry Genis — AlphaGeometry — A new approach to using LLMs to understand complex mathematical problems.
Boosted stories of the week
Medium has acknowledged the value these articles offer to readers and has approved them for Boost.
Giggles & Gigabytes: The AI Comedy Corner
“By far, the greatest danger of artificial intelligence is that people conclude too early that they understand it.” — Eliezer Yudkowsky
That’s all for now! We hope you found something interesting, or at least vaguely useful, in this newsletter. If you have any questions, ideas for future editions, or want to chat about the fascinating world of AI, then reach out to us in the comments below.
Until next time,
Jim and Priya