New Image & Speech capabilities

Advertise | Job board | Community | Investor + Founders connect

Hey folks, today we’ve got the next version of Midjourney for image creation, AssemblyAI’s new, super powerful speech recognition model, 1 of the big 4 (PwC) bringing AI to its lawyers and the UK pushing forward with AI investments.

Want to build projects and make content around AI? I’m launching a maker-in-residence program. Learn about it here and apply!

Let’s get to it.

🤌 Our Picks

Highlights if you've only got 2 minutes

1/ Midjourney v5 is released to level up your AI art skills

Remember that crazy period when some kind of generative art stuff came out and changed the world years ago 6 months ago, and then we all got distracted by language models? Just kidding, but hot on the heels of GPT-4 is Midjourney v5, and it’s equally as awesome. The thread linked above goes through a whole host of examples, and the results are stunning, albeit with subtle improvements over v4. It strikes me as better at capturing angular views of faces, is slightly less cartoonish, and captures the ‘vibe’ of the prompts more effectively. I’m keen to see assessments from the art experts out there!

Check out this video for how to get access.

2/ AssemblyAI’s speech recognition model is the most powerful one yet

Time for Shazaam to take a leaf out of AssemblyAI’s book and level up their ability to spot song lyrics in a noisy room… These guys are releasing Conformer-1, a speech recognition model that’s achieving near-human-level performance after being trained on 650,000 hours of audio data. That’s about an entire lifetime, or a whole day for all of the people that open this email added together… It has 43% fewer errors on noisy data than other publications, a 29% speed-up at inference, and actually uses novel tech. Check out some live examples on their website here.

3/ PwC bring generative AI to their lawyers

OpenAI dived straight in with their demonstration that GPT-4 is a significantly better lawyer (amongst other things of course), and PwC is standing ready to put that to the test. Teaming up with the OpenAI-backed startup Harvey (back to that GOD DAMN brilliant Suits reference again), they’re using ChatGPT-powered tech to bring efficiency to due diligence and regulatory compliance. With 4,000+ lawyers globally, PwC is making a big, commercial push to integrate generative AI into their business here.

Non-paywalled Bloomberg article here.

4/ The UK government throws a bit of money at the AI industry

It’s a warm and fuzzy feeling when people in power decide to take note of the importance of the nerdy work we all do - and the UK government seems to have done something along those lines in the Spring budget. With a number of AI-specific allocations of money, it’s a true demonstration of how AI is taking hold of the public discourse. My favourite bit is our very own Kaggle competition in the form of an annual £1m prize for AI research; there’s also £3b in additional funding for high-growth companies (read: tech?); and the building of a quantum computing hub just for good order.

Data chatbot for DTC brands - your own always-on, instant analyst (Sponsor)

Use Zenlytic's AI analyst to ask what's driving your store's acquisition, conversion, and retention.

It's like having a data analyst who never sleeps. Who can instantly drill into your brand's data. Who never forgets to update your dashboards.

See Zoë here

🛠️ Cool Tools

Product launches, updates and demos

• Chatfuel AI - Take customer communication to the next level. (link)

• Ask your PDF - Upload a PDF to chat with it. (link)

• Paper pages on the Hugging Face Hub - Discover models, datasets and spaces for a specific paper. (link)

• Collato - AI search for product teams. (link)

• Klarity - Chat with your document and discover detailed insights. (link)

• Auto transcribing is now faster and more accurate with Banner Bear + Whisper. (link)

• Smart Chat - Transform your Obsidian notes into interactive AI-powered conversations. (link)

• Darby Dashboards - Connect your APIs/data sources and create dynamic dashboards in seconds. (link)

• DocLime - Say goodbye to manual document search and get answers within seconds. (link)

• Contract Reader launches smart contract reviews powered by GPT-4. (link)

• Chatbot UI Pro - It’s an open source ChatGPT clone that you can run locally. (link)

• What does this code do - ChatGPT under the hood to explain any piece of code you don't understand. (link)

• Nabla Copilot - Superpowers for health providers. (link)

• Replit + GPT4 - Ready to go templates for Node.js and Python. Just create a Repl, place your API key, and run. (link)

• Chatgpt Asteriods - A game coded by ChatGPT. (link)

• I gave GPT-4 a budget of $100 and told it to make as much money as possible. (link)

• Cursor - An IDE built for programming alongside GPT-4. (link)

• ChatFriday - A minimalistic UI for ChatGPT. (link)

• Cal, the meeting scheduler, announces AI powered search to its docs. (link)

• Double - Start using GPT to automatically research your leads on the internet and answer questions. (link)

🤓 Miscellaneous

News, podcasts, videos, blogs etc

• GPT-4’s successes, and GPT-4’s failures. (link)

• Interview with OpenAI’s Greg Brockman: GPT-4 isn’t perfect, but neither are you. (link)

• GPT-4, Google adds AI to productivity apps, local language models. (link)

• Stripe and OpenAI collaborate to monetize OpenAI’s flagship products and enhance Stripe with GPT-4. (link)

• The Vesuvius Challenge - A machine learning and computer vision competition to read the Herculaneum Papyri with $250k in prizes. (link)

• Reid Hoffman wrote a book called Impromptu with Open AI’s GPT-4. (link)

• The multi-modal, multi-model, multi-everything future of AGI. (link)

• Apple engineers testing ChatGPT-style tech as Siri faces ‘clunky code’ and other hurdles. (link)

• LinkedIn is adding AI tools for generating profile copy and job descriptions. (link)

• Unscripted AI NPCs in a first-of-its-kind Unreal Engine demo. (link)

• Foundation Model Ops - Powering the next wave of generative AI apps. (link)

• OpenAI checked to see whether GPT-4 could take over the world. (link)

• Microsoft rations access to AI hardware for internal teams. (link)

• These new projects show just how much more powerful GPT-4 is. (link, without paywall here)

• Chat GPT4 - Is the world prepared for the coming AI storm? (link)

• Should we automate the CEO? (link)

🎓 Learn

How-to’s and resources

• Get structured output and clean dirty data with LLMs. (link)

• Building a SVB Q&A chatbot with Next.js and LangChain JS. (link)

• Create a basic 3D game prototype like Doom using GPT-4. (link)

👋 Too many links?! I created a database for all links mentioned in these emails. Refer 1 friend using this link and I'll send over the link database.

🔬 Research

Published research papers

• X-GPT from Microsoft - Connecting generalist X-Decoder with GPT-3. (link)

• Highly personalised text embedding for image manipulation by Stable Diffusion. (link)

• UPRISE from Microsoft - Universal prompt retrieval for improving zero-shot evaluation. (link)

• DeepMIM - Deep supervision for masked image modelling. (link)

• Re-ReND from Meta - Real-time rendering of NeRFs across devices. (link)

• Mesh Strikes Back - Fast and efficient human reconstruction from RGB videos. (link)

📰 Unclassifieds

Short, sponsored links 

• Alex Banks is hosting an emergency hackathon for the release of GPT-4 in London. (link)

• Anarchy on the couch - AI hackathon on 19th March at Celo HQ. (link)

• Lockheed Martin is hiring an AI/ML machine learning engineer. (link)

• Actalent is hiring a machine learning engineer. (link)

Have a product, service, job, event, newsletter, app, book, movie, tool, or anything you'd like to share with over 45k subscribers? 
Advertise | Job board | Community | Investor + Founders connect

🖼 AI Images of the day

Funny memes and pics from around the web

Send this to 1 AI-curious friend and receive my AI project tracker database! Use this link.