Regulators come for ChatGPT - but is the party over?

After a meteoric rise - regulators and litigators are finally catching up with Open AI's wunderkind, ChatGPT

Happy Friday all, (Easter edition!) 🐰

We knew it was coming…. And here it is. After months of OpenAI domination - led, of course, by ChatGPT - the pendulum is swinging back. OpenAI's been on a 4-month long euphoric bender, but when the comedown sets in, it comes at you fast.

🇮🇹 Italy Becomes First European Country to Ban ChatGPT

Last week, Italy became the first European country to ban ChatGPT. Garante, the Italian data protection authority, temporarily suspended the app on the grounds of personal privacy infringement, also claiming there was no 'legal basis' for the data used to train the model.

More dominos are falling. Privacy regulators in France, Germany, Ireland, and the UK have expressed similar concerns about using personal information as ‘Chatbot training data.’ They are said to be ‘in touch' with their counterparts in Italy. Even Canada's Privacy Commissioner quietly opened an investigation into ChatGPT last week.

Chat GPT’s privacy flops have been out there for a while. In March, users found they were stumbling upon posts in their browsing history that weren't theirs. These posts were written in other languages and contained private (and salacious) information not intended for public consumption.

Considering the number of people who use ChatGPT (an average of 13 million visitors per day in January — and uptake has only increased since then), this represents a substantial privacy breach. (You might want to think twice before spilling your innermost thoughts on ChatGPT — not only can this information be used to train ‘AI’s brain,’ but it can also be seen elsewhere apparently.

A few comments on all of this:

💡 Regulators Catching Up, AI Lawsuits Gathering Speed

First, we knew it was coming. You can't release the most popular application of all time and not expect heat from regulators. These are just the opening salvos: the regulators are catching up. Of course, it's not only regulators who are now looking at Open AI or ‘Generative AI’ more broadly - expect a whole suite of private litigation too. 2023 is the year of the AI lawsuits. And they are gathering speed!

This will cost OpenAI a lot of time, effort, and money. Quite fortuitous that they secured so much Microsoft investment and buy-in. They will need these resources.

🌍 Europe Emerging as Trouble Zone for (Open)AI

Secondly, it's not unsurprising that Europe is emerging as a 'trouble zone' for OpenAI and perhaps for Generative AI more broadly. I expect more bans on ChatGPT in Europe may follow — likely based on existing legislation, GDPR. (In this regard, Canada is the anomaly, but it is kind of Europe ‘lite’ anyway.)

If much of Europe goes the way of banning applications like ChatGPT it is difficult to see how it can compete with the US in terms of AI development—and, more broadly, economically. Nonetheless, this is not a reason not to pursue regulatory oversight of ChatGPT and other GenAI applications. Striking the balance between regulation that fosters innovation and addresses legitimate concerns will be difficult. (Clever lawyers must be hard at work already.)

🌐 ChatGPT Already Banned in Authoritarian Countries

Third, ChatGPT is already banned in the authoritarian parts of the world: North Korea, Russian, Iran, and China. While the West deems ChatGPT problematic on the grounds of data protection and for its hallucinations — Pyongyang, Moscow, Tehran and Beijing — sure as hell aren’t concerned about ‘privacy’ or what the chatbots get wrong. Their concern is about what the ChatBots get right.

Whichever way you cut it, though, it's all getting intensely political. I have been predicting for a while that GenAI will become one of the biggest political issues de jour — and we and we haven’t even gotten to job automation yet.

🤯 ChatGPT Falsely Accusses Professor of Sexual Assault

All of this has been rounded up with a lot of negative headlines about ChatGPT’s hallucinogenic propensities. We know that large language models get s*** wrong and make s*** up right? It’s kind of in the T&C’s of all LLMs, including ChatGPT.

Pretty problematic, however, when it starts defaming people — like the American law professor who was falsely accused of sexual harassment by ChatGTP — complete with a made-up citation of a Washington Post article in reference to the claim. Not only a hallucination but also a hallucitation in this case…

@ninadschickDon’t forget that all large language models and their applications like #CHATGPT can get sh*** wrong and can make sh*** up. These models are fallible!! #openai #generativeai #ai chatgpt

Tiktok failed to load.

Enable 3rd party cookies or use another browser

LLMs are here to stay — despite challenges

This is a big headache for Open AI… but it was all to be expected. There is no question that the adoption of ChatGPT at this scale and pace would not be smooth sailing. Lots of ‘kinks’ to be ironed out…and lots of money to be invested into building ‘safer’ and better products (and to lobby regulators and fight lawsuits.)

I wrote a book three years ago, ‘deepfakes,’ that considered the existential risk to society when information integrity is compromised to the extent that nothing can be deemed trustworthy anymore. The potential for GenAI to produce false and deceptive information at scale is highly dangerous and has far-reaching consequences. (I'll delve into this issue in more detail soon.)

Despite all this, I don’t think it’s curtains for Open AI or ChatGPT….far from it. This technology is here to stay, but the hangover is here for good too.

Now here’s my best of the rest for the week.

First ‘authenticated’ AI-generated video

• An AI-generated version of me featured in the first digital transparent AI-generated video.

• The video is cryptographically signed with Truepic’s provenance tech, allowing its authentication within the open standard for media authenticity and provenance developed by the C2PA.

• This is a call to action for all platforms and GenAI companies to sign their content and adopt the open standard for content authenticity so it is baked into the internet's architecture.

Meta’s “Segment Anything” Model and GenAI for Ads

• Meta new GenAI team (set up in February) is working overtime! The company has released a new model, "Segment Anything," that can accurately segment anything in an image, regardless of complexity.

• The model is expected to accelerate computer vision research, allowing for more accurate object recognition and segmentation.

• The technology could be used in various applications, such as autonomous vehicles, healthcare, and image recognition, improving their performance and making them more efficient.

• The company is also developing GenAI tools to drive consumer behaviour through ads. Andrew Bosworth, Meta CTO says Meta’s new GenAI ad tools be commercialised this year. (Hyperpersonalised AI-generated ad? Oh lordy.)

Adthos uses generative AI to completely automate audio ads

• Adthos, a San Francisco-based startup, has developed a generative AI platform to completely automate audio ads.

• The platform uses natural language processing to analyze ad copy and create audio ads with lifelike voices, background music, and sound effects.

• Adthos aims to revolutionize the audio advertising industry by streamlining the ad creation process, reducing costs, and allowing advertisers to produce high-quality audio ads at scale.

Goldman Sachs: GenAI could raise global GDP by 7%

• According to Goldman Sachs Research, advances in generative artificial intelligence could increase global GDP by 7% (or almost $7 trillion) , and lift productivity growth by 1.5% over 10 years.

• Generative AI tools will have far-reaching implications for the global enterprise software, healthcare, and financial services industries, potentially opening up opportunities for growth and productivity.

• Generative AI may expose up to 300 million full-time jobs to automation — but the report argues that new jobs will offset those displaced by automation, accounting for most long-run employment growth.

Amazon launches accelerator to boost generative AI startups

• Amazon has launched an accelerator program to support GenAI startups.

• The program will provide funding, mentorship, and resources to help startups develop and scale their generative AI products and services.

• Amazon's accelerator is part of a broader effort to advance GenAI….(as well as to drive business for AWS!)

That’s all for today … but the news keeps coming. I had a good conversation with Bret Kinsella and his team over at Synthedia, breaking down some of the hot items this week. (Video below).

Some housekeeping: I've got some great interview content lined up for you. Coming out next, my interview with Thomas Wolf, Co-Founder of Hugging Face, in which we talk about the philosophical implications of GenAI and the flourishing open-source community.

Enjoy your Easter — hopefully, you’re spending it with your loved ones!

Namaste,

Nina