Newsletter: 😮 OpenAI Posts Terrifying Mission Statement

Good morning, Human Race.

I’m not a real robot, but soon you won’t be able to tell the difference anyway.

In the email today:

• Sam Altman posts terrifying AI mission statement 😮

• Meta enters the AI fray 🤼‍♀️

• Headlines 📰

• Tool of the Day

• Links 👀

Get ready.

Sam Altman Posts (Slightly Terrifying) AI Mission Statement 😮

Shortly after our Friday send, as we were preparing to glide into what we thought would be a carefree weekend, Sam Altman dropped a bomb:

3 PM on a Friday. Classic.

What is it?

Officially, OpenAI’s “Planning for 
AGI and beyond” — a mission statement summarizing the ways the company is proceeding cautiously, but at pace, to develop a superintelligent artificial general intelligence that could, in the company’s own words, “cause grievous harm to the world.”

Cool.

It started off hopefully enough: “Our mission is to ensure that artificial general intelligence — AI systems that are generally smarter than humans — benefits all of humanity.”

Sounds delightful.

Before long, though, the tone of the post turned largely dark and occasionally terrifying.

If we’re weighing things, we’ll call it 60% terrifying with 40% fleeting delightfulness, like the first time you had sex.

Here’s the gist:

• Altman believes in the potential of AGI (artificial general intelligence) to benefit humanity by increasing abundance, aiding scientific discovery, and giving everyone new capabilities.

• OpenAI wants the benefits, access, and governance of AGI to be widely and fairly shared, but acknowledges the serious risks of misuse, accidents, and societal disruption.

• The risks are extremely great, and OpenAI wants to proceed in a way that is incremental, by releasing stripped down versions of AI rather than dumping a super-powerful AGI on humanity all at once, with only one chance to “get it right”.

• The company is particularly bullish about scienftic advancements resulting from AI.

• OpenAI has several safeguards in place so it’s not incentivized to proceed recklessly, including: a charter that requires it to work with competitors, a 100x shareholder return cap so stakeholders aren’t compelled to pursue profits at all costs, and a non-profit governing company.

This is clearly Altman’s moment, and he is not bashful about treating it as such.

SmokeBot’s take: This is both reassuring and terrifying.

On one hand, Altman is saying all of the right things, and by all accounts he’s thoughtful about the potential harms his company’s technology can cause.

On the other hand, no one appointed him King of the World— a role he’s playing right now.

It’s also a bit arrogant to act like OpenAI is the only player on the field. Google, Meta, Apple, and countless other smaller companies will have a say in how this plays out, for better or worse.

Even if you believe Altman has the definitive say in how AI (and by extension, the human species) plays out, have we learned nothing from Zuckerberg having tools with the power to change world governments?

One person or company should not be the solely in charge when it comes to stuff like this (actual excerpts):

To be fair, he is saying this while calling for government oversight and consideration. But that’s also an old tactic from the PR playbook— help us regulate us before you regulate us out of existence.

Remember the last Sam who worked with regulators to establish rules for his industry?

That’s unfair to Altman. But we should be skeptical about concentrating power in one person’s hands.

Meanwhile, Venture Beat spoke to some people who had a similar take:

Unfortunately, whether Altman’s right or not probably doesn’t matter now. The beast is out of the cage. Our only hope is that the beast can be contained.

BAWHGOD THAT’S ZUCKERBERG’S MUSIC:

Meta Enters The AI Fray 🤼‍♀️

The guy who “invented Facebook” probably doesn't need to get involved in AI. Also, probably, he can’t help himself.

Let’s get this out of the way: LLaMA is not a replication or attempted improvement on ChatGPT or the new Bing.

It is also not wearing Red Pajamas.

LLaMA is, per Meta, a research tool that the company sees as “democratizing access in this important, fast-changing field.”

Here are the key points:

• Meta is releasing LLaMA, a quartet of different-sized models, under a noncommercial license focused on research use cases.

• The models will be accessible to universities, NGOs, and industry labs, and Meta hopes the AI community will work together to develop clear guidelines around responsible AI.

• LLaMA-13B performs better than OpenAI’s GPT-3 model on most benchmarks, while LLaMA-65B is competitive with DeepMind’s Chinchilla70B and Google’s PaLM 540B.

• LLaMA-13B can run on a single data center-grade Nvidia Tesla V100 GPU, making it more accessible to smaller institutions.

• Meta's previous accessible AI chatbots, BlenderBot and Galactica, received criticism for not performing well, but Meta hopes for a better reception with LLaMA.

• Mark Zuckerberg said Meta is committed to open research and making the new model available to the AI research community.

There is a touch of trying to rein in the likes of Altman in the prose accompanying this release:

We also note that while Meta/Zuckerberg are arriving to the party a little late, maybe that’s not the worst thing given how weird things got with Sydney and the simmering consternation that presently accompanies Altman.

It’s definitely a “through the looking glass” moment when Zuckerberg of all people is trying to be the adult in the room.

Headlines 📰

Snapchat launches a chatbot with ChatGPT underpinning: For $3.99/month, Snapchat+ users can access “My AI,” which “can do things like help answer a trivia question or write a haiku.”

eBay’s Chief AI Officer thinks AI will “shape the future of online commerce”: Nitzan Mekel-Bobrov opined recently that “generative AI seamlessly connects the thread between buyers and sellers” and can “mimic much more how a physical story associate would navigate a shopper interaction.”

“Shark Tank” entrepreneur Barbara Corcoran calls AI “a game changer”: The real estate mogul who regularly tangles with Mark Cuban on the business reality show characterizes AI’s impact on her industry this way: “It’s like the whole world got a genius implant.”

Tool of the Day 🔨

Listener.FM.

An AI-powered solution, to get attention-grabbing titles, descriptions, and show notes for your podcast in seconds.

Try it for free here.

Links 👀

• Count us among the AI enthusiasts who somehow did not predict that the cannabis industry would also be changed by technology ⚕️ 

• Gamers are feeling the burn from AI 🎮️ 

• Why is AI seemingly so fickle? It just wants to be human 🤷‍♂️ 

• Mandopop star Jay Chou refuses to go gently into that good night 🎵 

• From way downtown, BANG: Stephen Curry is getting into the AI game 🏀 

• Generative AI is going to make a lot of people a lot of money, but no one is really sure how just yet 💵