←Ben’s Bites Waybackbensbites.beehiiv.com/p/editing-image-ease24 Jan 2023sent to 21,759opened 58.5%clicks 10,196unsubs 39← 23 Jan · 25 Jan → · random

Hey everyone. This is Ben's Bites, We're the Chipotle of AI: fast, fresh, and always on the cutting-edge. Plus we always give you a little extra 😉 

I built an app demo that lets you analyse and qualify leads734 using Zapier and OpenAI. You can clone/remix it for yourself. Essentially when someone fills out a form, AI will tell you whether its important, neutral or unimportant. This could be used for a gazillion use cases. If you build anything with Zapier + OpenAI, ping me @bentossell on Twitter!

Let's get to it.


Have a product, service, job, event, newsletter, app, book, movie, tool, or anything you'd like to share with over 20k subscribers? Check out our sponsor options225.

Join Discord93


🤌 Ben's Picks

1/

If you find it as painfully uncomfortable to listen back to your own voice recordings as I do, then it’s time for you to despair, because voice synthesis AI is starting to gain traction amongst investor circles. ElevenLabs received a $2m332 pre-seed for their company that produces realistic text-to-audio voice synthesis. They also extend existing recordings in audio-to-audio development, with long-term plans for this to work across languages. Yes, that’s right, your pain might one day be available in every language.

It’ll be pretty cool for the third stage in the big AI wave to begin with synthetic voice generation (the first being generative art, the second being chat-bots). With the accelerating popularity of audio-books, which are currently very expensive to produce, there could be some serious revenue in this realm with equally serious societal benefits.

And for a sneak peek at a real life pitch deck for a real life tech start-up with real life VC funding, this report442 includes the ElevenLabs one!

2/

The first bit of big gun shade has officially been thrown, with Meta’s Chief AI Scientist Yann LeCun saying that ChatGPT isn’t that innovative373, that lots of labs already do it, and Meta could and will do so too. Sounds a little bit sour, and reminds me of the Big Short quote “being late is the same as being wrong”. Yann, perhaps you’re just a little bit late? (To be fair, he does kind of have a point…)

3/

Is Artificial Domain Intelligence the route to commercialising AGI? I’ve been enjoying the pieces of discussion that have been circulating on domain specialisation using generative AI models, and Lead Data Scientist for FinCrime at KPMG summarises it well here427.

4/

Playground AI announces AI-first image editing. Instruct an AI to synthesise spectacular yet subtle edits. We're seeing a lot of products implementing InstructPix2Pix596 - a method for editing images from human instructions, which I believe is what's being used here. (link777)

🔥 Launched Today 🔥 (Sponsor)

Gallery of Automated Artistry is giving away their first shirt #0001/1825 on Twitter - potentially owning a piece of history sounds pretty cool.

The first-ever AI-generated high-end fashion brand in collaboration with top creatives (1.5M followers) 🖼️

Each design is serialized and only ever sold once.

They're giving away a piece of AI history, check out their first tweet. (link293)


🛠️ Cool Tools

  • Character AI - Generate realistic, intelligent, and interactive AI characters. (link916)

  • AI Tools FYI - Find amazing AI tools that make your life easy! (link924)

  • Stylized - Virtual product staging to create photos that sell. (link500)

  • Mad Genius - Directory of 500+ AI tools updated daily. (link857)

  • Algo - Conversational AI chatbot with less chatter and more control over personal data. (link311)

  • Cover Letter AI - Analyse your resume and write a cover letter that highlights info relevant to the job description. (link518)

  • Drayk.it - Make AI drake songs about anything. (link373)

  • Backend GPT - Using LLMs as backend. Scale AI hackathon’s 1st place project. (link461)

  • Rephraser - Text rephrasing tool using OpenAI's API. (link442)

  • Magician by Diagram - A design tool for Figma powered by AI is now in public beta. (link468)

  • Tome, the AI-powered storytelling format, has unveiled a multimodal vision for AI and released a new suite of generative storytelling features built for intuitive collaboration with AI, including text rewriting, length and tone adjustments, and generative prompt bar customisation. (link490†)


🤓 Miscellaneous

  • OpenAI and Microsoft extend partnership. (link173)

  • Alexa & ChatGPT - A match made in AI heaven. (link380)

  • The AI assist - A venture capitalists experience with AI as he comes back to coding. (link270)

  • Underfitted Talent Collective - A community of hand-curated AI professionals open to new opportunities. (link198)

  • The human-AI partnership - ChatGPT and Reid Hoffman talk about how AI amplifies human potential. (link190)

  • How Microsoft’s stumbles led to its OpenAI alliance. (link177)

  • The era of AI-first products. (link501†)


🎓 Learn

  • How ChatGPT actually works? (link661)

  • Build a game using ChatGPT and Replit. (link297)

👋 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

  • H3 - Generative language modelling with only 2 attention layers. (link110) One of the authors explains the model in this thread148.

  • Green hierarchical vision transformer for masked image modelling. (link95)

  • HexPlane: A fast representation for dynamic 3D scenes by 6 planes only. (link125)

  • Summarise the past to predict the future - Natural language descriptions of context can forecast the next object interactions. (link136)

  • BallGAN - 3D-aware image synthesis with a spherical background. (link113)

  • Is ChatGPT a good translator? Results: yes for high-resource European languages, not much for low resource or distant languages. (link134)

  • InfiniCity - Constructing and rendering an unconstrained, large and 3D-grounded environment from random noises. (paper85, demo116)

  • Zorro: The masked multimodal transformer. (link195)

  • LEGO-Net: Learning regular rearrangements of objects in rooms. (link118)

  • StyleGAN-T: Unlocking the power of GANs for fast, large scale text-to-image synthesis. (link299)

  • Batch prompting for efficient inference with LLM APIs. (link276)


📰 Unclassifieds

Have a product, service, job, event, newsletter, app, book, movie, tool, or anything you'd like to share with over 20k subscribers? Check out our sponsor options225.


🖼 AI images of the day


Editing image with ease
PLUS: "ChatGPT isn't that innovative" from Meta's Chief AI Scientist

less more clicks · number = clicks † = site gone, link opens the Wayback Machine

Most clicked in this issue

924aitools.fyi
916mycharacter.ai
857madgenius.co
777twitter.com/Suhail/status/1617651599710752768…
734twitter.com/bentossell/status/161758001189631…
661youtube.com/watch?v=x_bw_IHjCWU
596timothybrooks.com/instruct-pix2pix
518coverletterai.app
501moreentropy.com/about
500stylized.ai
490beta.tome.app
468magician.design

2,974 readers clicked at least once. 48 links, 48 with clicks. Heat is relative to the most clicked link in this issue.

Beehiiv · digest