Hard Numbers: Profitable prompts, Happy birthday ChatGPT, AI goes superhuman, Office chatbots, Self-dealing at OpenAI, Saying Oui to Mistral

$200,000: Want an image of a dog? DALL-E could spit out any breed. Want an Australian shepherd with a blue merle coat and heterochromia in front of a backdrop of lush, green hills? Now you’re starting to write like a prompt engineer, and that could be lucrative. Companies are paying up to $200,000 for full-time AI “prompt engineering” roles, placing a premium on this newfangled skill. It's all about descriptive fine-tuning of language to get desired results.

1: Can you believe it’s only been one year since ChatGPT launched? It all started when OpenAI CEO Sam Altman tweeted, “today we launched ChatGPT. Try talking with it here.” Since then, the chatbot has claimed hundreds of millions of users.

56: Skynet, anyone? No thanks, say 56% of Americans, who are concerned with AI gaining “superhuman capabilities” and support policies to prevent it, according to a new poll by the AI Policy Institute.

$51 million: In 2019, OpenAI reportedly agreed to buy $51 million worth of chips from Rain, a “neuromorphic” chip-making startup, meant to mirror the activity of the human brain. Why is this making news now? According to Wired, OpenAI’s Sam Altman personally invested $1 million in the company.

$20: You work at a big company and need help sifting through sprawling databases for a single piece of information. Enter AI. Amazon’s new chatbot, called Q, costs $20 a month and aims to help with tasks like “summarizing strategy documents, filling out internal support tickets, and answering questions about company policy.” It’s Amazon’s answer to Microsoft’s work chatbot, Copilot, released in September.

$2 billion: French AI startup Mistral is about to close a new funding round that would value it at $2 billion. The new round, worth $487 million, includes investment from venture capital giant Andreessen Horowitz, along with chipmaker NVIDIA and the business software firm Salesforce. Mistral, founded less than a year ago, boasts an open-source large language model that it hopes will rival OpenAI’s (ironically) closed-source model, GPT4. What’s the difference? Open-source LLMs publish their source code so it can be studied and third-party developers can build off of it.