The inside story of ChatGPT

Hey! this is Ben's Bites...dishing up our daily AI news with a side of excessive tabs, all washed down with a drop in productivity.

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 options.

Join Discord

🤌 Ben's Picks

1/

The inside story of ChatGPT: How OpenAI founder Sam Altman built the world’s hottest technology with billions from Microsoft. (long read, but worth it)

This article discusses the impact of OpenAI's ChatGPT chatbot, which has been used to debug code, write recipes, scripts, and more, and how it has sparked a revolution.

2/

Five pieces of advice for those building in AI right now.

I really like Nathan and the way he articulates what he's thinking when building products. I also happen to really like the products he builds. Some advice he shares in his thread: Resist broad generalisations, AI isn't anyones moat, products using AI aren't just 'wrappers', ignore the hype, and AI-powered applications are mostly not about AI. (link)

3/

After inking its OpenAI deal, Shutterstock rolls out a generative AI toolkit to create images based on text prompts.

If stock photography firm Shutterstock were a person in the 1960s, they’d be a long-bearded hippie singing “Make love, not war man” from the rooftops.

Indeed, they’re proving to be a shining example of a large company embracing AI with strength. Whilst Getty Images have been getting into fisty-cuffs and legal battles over rights to use their images for AI art model training, Shutterstock have been handing their images to OpenAI and others for training, and training their own models, and giving licensing rights to the artistic creators. They’re now in a strong position to benefit from the tech and be on the right side of its users.

4/

Modeling a robot with some added difficulty.

Are robots gonna take over the world? Maybe. Probably not. Should we spend 4-5 hours using AI to design awesome-looking ones with little antennas and an industrial aesthetic? Absolutely.

This post modelling robots is a great example of the effort that is still required to produce truly excellent results: artistry, patience, and skill. It works through the creation of game robots applying prompting, modelling, texture projection and more, taking 4-5 hours. And this effort is a goodsign. It’s living proof that AI is not here to imminently steal creative and process-driven jobs, but to help enhance our roles in them, increase productivity, and improve output.

If stock photography firm Shutterstock were a person in the 1960s, they’d be a long-bearded hippie singing “Make love, not war man” from the rooftops.

🛠️ Cool Tools

• Gift Assistant - AI powered gift ideas. (link)

• TRL 0.2.0 - A library to train language models with reinforcement learning. (link)

• Andi - Smart search assistant for the next generation. (link)

• OpenJourney: Midjourney, but open source. (link)

• NeevaAI officially launched on Product Hunt today. NeevaAI allows users to get a single coherent answer to their queries while also showing citation cards. (link)

• HappyMama - Get instant answers to your pregnancy questions. (link)

• Mini Yohei 2.0 - A VC expert who answers questions using knowledge from 100+ VC blogs. (link)

🤓 Miscellaneous

• Atomic AI raises $35 million in series A for AI-driven RNA drug discovery. (link)

• Mosaic claims training Stable Diffusion from scratch costs less than $160k. (link)

• Generative AI won’t revolutionise game development just yet. (link)

• Amazon begs employees not to leak corporate secrets to ChatGPT. (link)

• ChatGPT explains how to invest during a recession - a 5-part strategy for balancing a portfolio. (link)

• ChatGPT creates a new ChatGPT. (link)

• This startup is using AI to unearth new smells. (link)

• AGI is Silicon Valley’s first native cryptid. (link)

• Replacing a SQL analyst with 26 recursive GPT prompts. (link)

• CNET is testing an AI engine. Here's what we've learned, mistakes and all. (link)

🎓 Learn

• Getting started with LLMs using LangChain. (link)

• Deploying AI-driven apps on Vercel. (link)

• Introduction to embeddings - The bread and butter of language models. (link)

🔬 Research

• Open source ViT-G/14 CLIP model by LAION achieves 80.1% zero-shot accuracy. (link)

• Scaling unlocks emergent abilities in language models. (link)

• Learning with quired hints for decision-making problems under uncertainty. (link)

• PushWorld: A benchmark for manipulation planning with tools and movable obstacles. (link)

• Interactive-chain prompting to reduce translation ambiguities and improve translation quality. (link)

• A holistic cascade system, benchmark, and human evaluation protocol for expressive speech-to-speech translation. (link)

• A multilingual, multispeaker speech synthesis system which can generate speech with the proper accent. (link)

• XLM-V - Overcoming the vocabulary bottleneck in multilingual masked language models. (link)

• Pre-computed memory or on-the-fly encoding? A hybrid approach to retrieval augmentation. (link)

• ClimaX - A foundation model for weather and climate. (link)

• Separate and diffuse - Using a pre-trained diffusion model for improving source separation. (link)

• Imitating human behaviour with diffusion models. (link)

📰 Unclassifieds

• Ben's Bites is hiring! I'm on the lookout for someone to own growth. It could be setting up meme marketing, side projects, testing ads, running a student-ambassador program, (almost) anything is on the table. If you want to join me and grow this thing, email me here!

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 options.

🖼 AI images of the day

🤗 Share Ben's Bites

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

or copy/paste this link: https://bensbites.beehiiv.com/subscribe?ref=PLACEHOLDER

⭐️ How did we do?

⭐️ REAL REVIEWS