🍕 Sci-fi isn’t Sci-fi anymore

Hey there,

my head hurts for how much stuff happened this week. This edition will be fun and PACKED, so I hope you’re sitting down 😂

P.S.: there’s ONE last seat available for our Master in Prompt Engineering. You should take it.

This week’s news:

• Sci-fi costs 2 McDonald’s meals per month

• Pizza Bytes 🍕: Meta’s new AI experience, CIA’s AI tool, OpenAI is facing another lawsuit

Now let’s get started 🕺

Sci-fi costs 2 McDonald’s meals per month

OpenAI has announced the biggest overhaul to their most successful product: ChatGPT will be able to see, speak, and access the internet within two weeks.

What does it mean? I always teach my students to see AI as an assembly line where data can get into models, and some new data comes out of them. So far ChatGPT has been limited to one kind of data: text (+ voice on the iOS app), and the output has always been just text. Now we’ll be able to input text, voice, and images, and get text, audio, and images back, and ChatGPT will autonomously decide whether to add internet-sourced data too.

This sounds very abstract, so let’s talk about a concrete use case: in the website demo, OpenAI showed a user asking ChatGPT how to adjust the saddle of a bike, attaching a picture of the bike too. ChatGPT guided the user to the right procedure and even asked for a picture of the available tools to give further indications.

A friend of mine commented on LinkedIn saying “Can we just stop calling it ChatGPT and start calling this “JARVIS” instead?” (JARVIS is Iron Man’s assistant).

Behind that funny joke, there’s a whole world of considerations to do. First of all: sci-fi is literally not sci-fi anymore. It’s a 20$/month subscription. Think about this: the only thing stopping you from living in the future is the cost of a couple of McDonald's meals per month.

Let’s add another element to the mix: Microsoft last week announced Copilot, "your everyday AI companion". Copilot is an update of Windows 11 that will basically bring GPT-4 + Bing on every PC. It can access the web AND your local files, it's integrated with lots of Microsoft tools like Outlook, Edge, Snipping tool, and even Paint. It can also use the power of DALL-E 3 to generate images (which was also announced last week).

So basically in a single week, OpenAI has announced a Jarvis-like new experience, and Microsoft is the first to integrate that at the operative system level in everything we do.

Again, all of this happened in one week.

My head hurts.

Now, for sure there are thousands of issues that need to be ironed out: the privacy constraints are not trivial, the tools won’t work 100% for now, the use cases aren’t well defined, etc. etc.. That’s all true, and I generally don’t believe in anything I haven’t tried myself.

However, my goal with this newsletter is to get you thinking about something: one year ago, ChatGPT didn’t exist. One week ago, ChatGPT couldn’t see, speak, and access the internet. Where will we be in 5 years?

Pizza Bytes 🍕

• Spotify is using OpenAI's voice generation tech to pilot Voice Translation in podcasts, maintaining the speaker's style while translating episodes into different languages.

• Meta introduced new AI experiences that enhance connection and creativity. Users can now generate customized stickers, transform images with new features like Restyle and Backdrop, and interact with Meta AI, a personalized assistant available on messaging platforms and smart glasses.

• YouTube has introduced AI-powered tools for creators, including Dream Screen, which lets users add AI-generated video or image backgrounds to their videos. They have also added tools for brainstorming, drafting outlines, searching for music, and dubbing videos in different languages.

• The CIA is launching an AI-powered tool similar to OpenAI's ChatGPT to help analysts have better access to open-source intelligence. The tool will be used across the US intelligence community, allowing analysts to ask questions and receive answers from the AI program.

• SAP is integrating its AI tool, Joule, into its cloud enterprise portfolio, offering proactive and contextual insights across their solutions, enabling employees to ask questions in plain language and receive answers from extensive business data.

• WHOOP, a wearable fitness tech company, now has an AI-powered feature called WHOOP Coach. It gives personalized health and fitness tips based on biometric data, like sleep, workouts, and nutrition, using OpenAI's generative AI.

• NASA's Perseverance rover is using AI to train its robotic arm to identify and analyze rock and landscape features on Mars. This AI system could eventually help in the search for signs of life on other planets.

• French AI startup Mistral has released its Mistral 7B model, a high-performing language model that is free to use without restrictions. The model aims to support the open generative AI community.

• Kneron, a semiconductor startup, has raised an additional $49 million in funding to support the commercialization of its AI chips. The company aims to challenge Nvidia's dominance by designing chips that enable on-device AI, particularly for autonomous driving.

• OpenAI is facing yet another lawsuit, this time from notable authors and the Authors Guild, accusing the company of copyright infringement. The plaintiffs claim that OpenAI used their books to train its language models without permission or compensation.

• AI-generated naked images of young girls in Almendralejo, Spain, have been circulating on social media, involving the use of fully clothed photos and an app that creates nude images.

You reached the end of this edition 😢

I’ll talk to you next week.

Ciao 👋