#17 Google 2.0: Search Reimagined
Google announced a slew of "AI-upgrades" to existing suite of products & services including "search", Meta released an AI model for 6 data types, and Anthropic expanded context length to 100K tokens!
If you are new to Discovery Unlocked, here is our one-line pitch-
We talk about the most exciting 🛠️ tools/launches, 🐣 thoughts and 📑 developments of the week, on Friday (or sometimes saturday🙊).
But only one email per week, so no “AI fatigue”.
Here’s what the future of Google search is going to look like..
In store this week
🛠️ Tools
Boring Report- an app that uses AI to remove sensationalism from news [link]
Puddl- track OpenAI costs for free [link]
Schrodi- create engaging and personalized stories for children [link]
Flow- an AI-driven financial model for personal finance [link]
Feathery AI- Create Web Forms 10X faster with AI [link]
Scispace- AI research assistant [link]
📑 Developments
Google launches PaLM 2, a next-gen LLM, showcases Gemini, a multi-modal model, and makes Bard available for everyone at Google I/O ’23 [link]
Anthropic expands context length in Claude to 100K tokens [link]
Meta open-sources ImageBind, a multisensory AI model that combines six types of data [link]
HuggingFace releases Transformers Agents [link]
Stability AI releases Stable Animation SDK, a powerful text-to-animation tool for developers [link]
IBM launches Watsonx, an AI platform for enterprises [link]
Scale AI launches Donovan and EGP, two new platforms for defense and enterprise AI [link]
Assembly AI introduces LeMUR, a new framework for applying powerful LLMs to transcribed speech [link]
Airtable AI to deploy AI-powered applications for the enterprise [link]
Wendy’s x Google- AI chatbot that takes drive-thru orders [link]
Microsoft working with AMD on expansion into AI processors [link]
+ 🐣 Thoughts
The debates inside every VC firm, Warren Buffet on AI, the Future of tech might be close to invisible & more,
+ 🥷 Random
Outcome of Rewind’s unusual fundraise, BVP’s $1Bn AI Fund, List of OpenLLMs, Responsible AI startups, and How to get smarter results from GPT 4 yourself.
🛠️ Tools
Boring Report- an app that uses AI to remove sensationalism from news [link]
Puddl- track OpenAI costs for free [link]
Schrodi- create engaging and personalized stories for children [link]
Flow- an AI-driven financial model for personal finance [link]
Feathery AI- Create Web Forms 10X faster with AI [link]
Feathery's AI form assistant lets you create high-quality forms in just a few seconds. Polish the final experience in our powerful editor.
Scispace- AI research assistant [link]
Highlight confusing text, math, and tables to get a simple explanation or ask follow-up questions and get instant answers.
🐣 Thoughts
The debate happening inside of every VC firm
AI- Invest now vs sit out and watch
Read here [link]
Image Courtesy- TechCrunch+ Warren Buffet’s take on AI
Warren Buffett voiced both concern and amazement over AI. He also expressed his worries, comparing AI to an atom bomb: “I know we won’t be able to un-invent it and, you know, we did invent, for very, very good reason, the atom bomb in World War II.”What does it take to be an AI founder? [Podcast]
Sharif Shameem of Lexica shares his views on building in the AI-age.
Humane’s first-ever public demo
Screenless-Seamless-Sensing - The Future of technology might almost be invisible- Imran Chaudhary, CEO of HumaneThe Ask Jeeves-ification of online search
Ask Jeeves launched in 1997, and the idea was that you’d type a natural-language query into the box, and the valet would come back with an answer.
Read the take here [link]
📑 Developments
Google launches PaLM 2, a next-gen LLM, showcases Gemini, a multi-modal model, and makes Bard available for everyone at Google I/O ’23
Here’s a snapshot of all AI related annoucements made-
0:28 Google Photos (Magic Eraser)
1:02 Google Maps (Immersive View for Routes)
1:10 PaLM 2
1:23 Sec-PaLM
1:29 Med-PaLM 2
1:42 Bard
2:40 Generative AI in Google Sheets and Docs
3:30 Generative AI in Google Search
4:32 Generative AI in Google Cloud
4:39 Vertex AI
4:58 Project Tailwind (AI First Notebook)
5:30 Responsible AIRead more about entire set of launches here [link]
Anthropic expands context length in Claude to 100K tokens
Anthropic has expanded Claude’s context window to 100,000 tokens of text, corresponding to around 75K words. Users can now submit hundreds of pages of materials for Claude to digest and analyze, and conversations with Claude can go on for hours or days.
Read more here [link]
Meta open-sources ImageBind, a multisensory AI model that combines six types of data
The new open-source AI model links together multiple streams of data, including text, audio, visual data, temperature, and movement readings. The model learns a single embedding, or shared representation space, not just for text, image/video, and audio, but also for sensors that record depth (3D), thermal (infrared radiation), and inertial measurement units (IMU), which calculate motion and position. ImageBind equips machines with a holistic understanding that connects objects in a photo with how they will sound, their 3D shape, how warm or cold they are, and how they move.
Read more here [link]
HuggingFace releases Transformers Agents
Create an agent using LLMs (OpenAssistant, StarCoder, OpenAI etc.) and start talking to transformers and diffusers. It responds to complex queries and offers a chat mode. Create images using your words, have the agent read the summary of websites out loud, and read through a PDF.
Read more here [link]
Stability AI releases Stable Animation SDK, a powerful text-to-animation tool for developers
Stable Animation SDK supports
- Text to animation: Animation is solely based on the text prompt
- Text input + initial image input: Users provide an image to base their animation on
- Input video + text input: Users provide an initial video to base their animation on
Read more here [link]IBM launches Watsonx, an AI platform for enterprises
With watsonx.ai, businesses can train, validate, tune, and deploy models with ease and build AI applications in a fraction of the time with a fraction of the data.
Read more here [link]
Scale AI launches Donovan and EGP, two new platforms for defense and enterprise AI
Scale Donovan is an AI-powered decision-making platform that helps defense operators understand, plan, and act in minutes instead of weeks. Donovan ingests and understands vast amounts of structured and unstructured data to enable operators and analysts to make sense of any aspect of the real world in minutes using simple natural language.
Scale EGI is an Enterprise Generative AI Platform that lets enterprises build and deploy generative AI applications for their business.More details here [link]
Assembly AI introduces LeMUR, a new framework for applying powerful LLMs to transcribed speech
LeMUR can quickly process audio transcripts for up to 10 hours worth of audio content which effectively translates into ~150K tokens. By contrast, off-the-shelf, common LLMs are only able to fit up to 8K or ~45 minutes worth of transcribed audio within their context window limits.Find details here [link]
Airtable AI to deploy AI-powered applications for the enterprise
Airtable’s new AI components and intuitive no-code interface make it simple for teams to integrate powerful AI capabilities into their own data and workflows. Organizations can power a wide range of processes all in one place – from producing job requirements, to managing marketing campaigns, to planning new research and product development initiatives.
Read more here [link]
Wendy’s x Google- AI chatbot that takes drive-thru orders
The fast-food chain is partnering with Google to train an AI model that knows what ‘biggie bags’ and ‘JBC’ mean.
Read more here [link]
Microsoft working with AMD on expansion into AI processors
Microsoft is working with Advanced Micro Devices Inc. (AMD) on the chipmaker’s expansion into artificial intelligence processors as part of a multi-pronged strategy to secure more of the highly coveted components. The companies are teaming up to offer an alternative to Nvidia, which dominates the market for AI-capable chips called graphics processing units.
Read details here [link]
🥷 Random
Rewind’s unusual fundraise sees committed offers from 170+ VCs
Rewind has decided to finalize the fundraise with NEA at an estimated valuation of $350 million, which is 4.5x of a16z’s valuation in the last round.
‘Responsible AI Challenge’ top 10 finalists
The finalist selection process was made possible by a panel of select individuals including AI academics, developers, and entrepreneurs. These include an app for gamified local learning, a bot to empower women in communities plagued by violence, and SaaS for confidential deployment solutions for LLMs among others.Find the finalists here [link]
Bessemer Venture Partners is committing $1 billion to invest in AI startups
Read details here [link]
Stack Overflow is ChatGPT Casualty: Traffic Down 14% in March
Read the detailed analysis here [link]
SmartGPT- How to get smarter results from GPT 4 yourself
SmartGPT, a system with evidence, should serve as your ultimate guide for boosting the automatic technical performance of GPT 4, without needing few shot exemplars.List of Open LLMs
A list of open-source LLMs available for commercial use.
Find the entire list here [link]
We hope you liked what we have featured. As always we cannot stress more how eager we are to hear from you. Your remarks, insights, and feedback would be super helpful, so please tell us what you liked, or what you didn’t.
Please, do tell your friends about us.
Here’s wishing you all a happy weekend.
See you again, next week.