Breaking AI News: Stay Ahead of the Curve with the Latest Tools and Resources & Generative AI Meets Biology: The Next Frontier in Life Sciences in protein design and discovery?
Your Weekly Shortcut to Deeptech Investing—Exclusive Trends & Startup Reports for VCs & Angels in Just 5 Minutes
This week…
1️⃣The latest AI news, startups and tools to try.
2️⃣Generative AI in Biology.
Breaking AI News: Stay Ahead of the Curve with the Latest Tools and Resources
By Lirone Samoun, Deeptech Expert
![[webp-to-jpg output image]](https://substackcdn.com/image/fetch/$s_!hyKO!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F79d7abdf-c760-43ff-9007-1d870042ecd8_1024x1024.jpeg "[webp-to-jpg output image]")
Today, I offer you a new article format. As, the AI world is moving very fast and especially, the last few weeks, I would like to give you an overview about what happened recently, some interesting start-up along with some cool tools that you could try.
❤️ If you like this format, please like this post and I would definitely make more similar posts in the future.
🔎 Latest news
Microsoft announces new Bing and Edge browser powered by upgraded ChatGPT AI
Microsoft integrates ChatGPT into Bing and Edge to revolutionize the Web. They would like to use conversational AI to create a new way to browse the web. Users will be able to chat to Bing like ChatGPT, asking questions and receiving answers in natural language. Microsoft said that this artificial intelligence is based on "the next generation of OpenAI's large language model, customized specifically for search." While the name GPT-4 was not mentioned, it was hinted at several times during the conference. And as if to confirm it, "it's much more powerful than ChatGPT"Runway, the original startup behind Stable Diffusion, has launched a generative AI for video
Use words and images to generate new videos out of existing ones. Gen-1 is a new Diffusion-based that allows you to realistically and consistently synthesize new videos by applying the composition and style of an image or text prompt to the structure of your source video. Check it out. You can even signup for an early access.Remember Meta’s new text-to-video AI generator Make-A-Video (like DALL-E but for video) ? It looks like the generation of video and everything around it will take off this year.
Google announces Bard, a ChatGPT competitor powered by LaMDA
Google CEO Sundar Pichai revealed Bard, an experimental conversational AI service, powered by LaMDA, to compete with OpenAI’s ChatGPT. Google is opening it up to trusted testers ahead of making it more widely available to the public in the coming weeks.
OpenAI, Microsoft, Google, Amazon, Baidu (Chinese internet search major) that is planning to launch its own chatGPT alternative…The war of conversational AI is starting.
This is going to be exciting.
Google announces MusicML, a very impressive text-to-music model
MusicML is a language model that can turn descriptive text about music into actual music! Detailed in an academic paper, MusicLM was trained on a dataset of 280,000 hours of music to learn to generate coherent songs for descriptions of — as the creators put it — “significant complexity”
In 2023-2024, we will see tremendous effort about generative AI and this time, it’s not about video or image generation, but music generation.
![[video-to-gif output image]](https://substackcdn.com/image/fetch/$s_!hOjk!,w_1456,c_limit,f_auto,q_auto:good,fl_lossy/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F701ee324-5c71-4111-bc6d-48f88f2eee17_600x338.gif "[video-to-gif output image]")
The Multi-Speaker Speech Wizard
Google Research has proposed SPEAR-TTS, a text-to-speech model that can convert written text into spoken words with natural-sounding voices, even for multiple speakers, with very little supervision needed. This allows the system to learn from audio recordings of people speaking, rather than needing a lot of text-to-speech examples.Galileo AI announces a generative AI for User Interface design. It can turn a simple text description into high-fidelity editable UI designs. Signup here: https://useGalileo.ai
![[video-to-gif output image]](https://substackcdn.com/image/fetch/$s_!hOjk!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F701ee324-5c71-4111-bc6d-48f88f2eee17_600x338.gif "[video-to-gif output image]")
![[video-to-gif output image]](https://substackcdn.com/image/fetch/$s_!9rUk!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fdc4973f5-d8eb-4d26-a5a9-cbdc4ee9334f_600x600.gif "[video-to-gif output image]")
Google releases MetNet-2, a deep learning model that can predict rain up to 12 hours in advance.
The premiere of the streaming app of the NBA is totally INSANE. A mix of #AI, CG and much more. Check it out.
Semantics-guided natural synthesis 👉Alibaba AI unveils a novel semantics-guided synthesis of natural scenes.
![[video-to-gif output image]](https://substackcdn.com/image/fetch/$s_!0C9y!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fa5b139af-6f51-4ca2-94be-f5fe72f67e40_600x338.gif "[video-to-gif output image]")
Ex-Tesla Director of AI Andrej Karpathy will be joining back OpenAI. The company is currently on a hiring spree.
🔬Research
Text-driven visual synthesis with latent diffusion prior. (link)
Exploring the limits of ChatGPT for query or aspect-based text summarization. (link)
Pix2pix 3D - 3D aware conditional image synthesis. (link)
PRedItOR from Adobe - Text-guided image editing with diffusion prior. (link)
💻 CooI AI Tools / Startup
Research GPT - An open-source research assistant that allows you to have a conversation with a research paper. The app extracts text from a PDF and creates embeddings to generate a response to a user's question using the OpenAI API.
GPT Playground - Use GPT-3 with more control and presets
Trav - Create your next travel adventure.
Fireflies.ai - Help your team record, transcribe, search, and analyse meeting notes.
Piano Genie - Have some fun pretending you're a piano virtuoso using machine learning.
Cognify AI studio - Transform your photos into stunning designs with the power of AI.
ArxivGPT - Summarize an arXiv paper and provide key insights.
Murf AI - AI Voice Generator: Versatile Text to Speech Software
Jasper - AI Copywriting & Content Generation for Teams
Supermeme : Generate Memes with AI
Poised :AI-powered communication coach that helps you speak with confidence and clarity
Synthesia: Create videos from plain text in minutes
👋That’is for this edition.
❤️Again, If you like this kind of format, please like this post, comment it to show your interest and I would definitely make more similar posts in the future.
I have plenty of cool AI tools, start up to show you =D
Generative AI Meets Biology: The Next Frontier in Life Sciences in protein design and discovery?
By Suraj Nair, External Contributor
Generative AI in Biology- Designing proteins from scratch for specific use cases in drug discovery and diagnostics.
What is going on?
AlphaFold from DeepMind and RoseTTAFold from Institute for Protein Design (IPD) at the University of Washington were among the first few algorithms to solve the first part of the problem: deducing protein structure from an amino acid sequence. AlphaFold2 currently has predicted 200 million protein structures with very high accuracy. In November 2022, Meta AI released the structures of more than 600 million proteins in the ESM Metagenomics Atlas database. Meta’s approach uses generative language models to generate its structures while AlphaFold, on the other hand, uses pattern recognition tools such as transformers to identify interactions between amino acids at long distances in protein structures.
Solving for the second part of the problem which is essentially protein functionality, Nvidia along with Evozyne announced the development of two new proteins for specific functionalities, namely to reduce CO2 and to cure congenital diseases. It followed the announcement by Absci who developed a zero-shot generative AI model to design novel antibodies in E. Coli, which means the model designed new protein molecules which had not been observed during its training. Similarly, Insilico Medicine used a combination of its Pharma.AI platform along with AlphaFold to generate a new drug candidate for primary liver cancer.
What does it mean?
One of the key challenges in the use of computational tools in biology is the inherent complexity of the systems in question. Take for example, proteins: they have a specific 3D structure/folding which gives them their functionality. There are two major problems in the modern study of proteins: predicting the structure of proteins from the underlying amino acid sequence, and secondly associating the structure to the right functionality. To solve these problems, large data sets are required which can be used to train specialized algorithms and generate predictive outcomes which are validated through wet lab experiments. The experimental data is fed back into the algorithms to improve their performance. This approach, coupled with sophisticated computational techniques, has shown promise and proved to be highly successful in the recent past.
The second part of the problem, namely associating a protein’s functionality to its structure, is significantly more difficult. Built on training through large data sets on protein structure-function relations, generative AI has the ability to design a novel protein with the desired application from scratch, thereby addressing the protein functionality problem.
Why does it matter?
💸For markets: new business opportunities.
Experts predict generative AI could generate more than $1 trillion in value for the healthcare industry by 2040 with use cases in drug discovery, diagnosis, personalized medicine, clinical trials and many more. Several companies have raised significant amounts of venture capital funding to develop advanced algorithms for biology applications. Some of the companies in this sector include Insilico Medicine, Evozyne, Generate Biomedicines, IBM, Absci who are developing advanced generative algorithms for protein design and discovery applications. A very low percentage of the drug candidates that go through the FDA approval process are ultimately approved for clinical use, but AI models have the potential to find more biologically relevant compounds in significantly lesser time periods with better chances of approval.
🧑🏿🤝🧑🏻For society: reduce the cost of diagnosis and therapy.
Use of generative AI can significantly reduce the cost of protein design and discovery by reducing the number of experiments performed. This can also reduce the timelines for developing a new molecule thereby lowering the cost of the diagnostic and therapeutic products and increasing their access to the population at large.
🔮What’s next?
The ability to design proteins with specific functionality can open up huge opportunities in drug discovery as well as diagnostics. Use of generative AI can reduce the timelines for new drug discovery and reduce the cost of manufacturing as well. It can also open the field of precision medicine in the future. As the technology matures and specific applications in biology are explored further, generative AI applications in biology are expected to attract significant venture capital investment. The science of proteins marks one of the frontiers of modern biology; the use of computational tools not only promises a variety of applications, but also gives us a clearer understanding of the fundamental nature of proteins.
Note: Read more about Ankur Capital and our work in deep science technologies here
➡️ Ankur Capital - Techsprouts
Very Insightful of Saying the impact of generative AI in Biotech, Great read, Suraj Nair.