💻 How AI Will Change Your Work Life

Big week in AI last week.

We covered GPT-4 from every angle, but that was only part of the story. Microsoft announced its AI copilot - GPT-4 technology built directly into Office 365 - and OpenAI CEO Sam Altman sat down with ABC News to talk about it all.

In the email today:

• Sam Altman’s ABC News interview 📺

• Microsoft adds AI to Office365 💻

• Headlines 📰

• Links 👀

Here we go.

Sam Altman’s ABC News Interview Didn’t Make Us Feel Any Better About Where AI Is Headed 📺

OpenAI CEO Sam Altman recently sat down with Rebecca Jarvis of ABC News to (we think) try to assuage an increasingly skeptical and nervous public about AI generally and ChatGPT/GPT-4 specifically. You can watch the video here.

Did he succeed? To our minds, not really, no.

Besides all of the obvious and helpful use cases of ChatGPT, there is genuine, sincere concern about what AI in general could become. Altman is simultaneously pressing forward and calling out the potential dangers of this technology.

SmokeBot sat through the entire interview so you don’t have to. Here are his key quotes and observations:

• “This will be the greatest technology humanity has yet developed.”

• “We’ve got to be cautious here…make our mistakes while the stakes are low.”

• Altman argued that releasing ChatGPT and then GPT-4 gradually to an unprepared world populace was definitely better than continuing to work on the technology in secret, then dropping an advanced, world-shaping technology all at once.

• Aside from “we are a little bit scared of this,” which is obviously the most terrifying thing Altman said, this quote also gave us pause: “The one thing I would like to see the government do is really come up to speed on understanding what’s happening…where our capabilities are.”

Carved into Altman’s Q&A with Jarvis was a shorter interview segment with OpenAI’s Chief Technology Officer, Mina Murati. She wasn’t exactly comic relief.

Murati conceded that even with GPT-4, “you could guide it to negative outcomes.” She contended that machines are still “easier to predict” than humans because “there’s a scientific process.” But in the same breath, she allowed that a “point of no return” could not be ruled out.

We have often praised Altman for his obvious brilliance and his openness. Additionally, he created this technology, and as such, he may not be the leader we elected but he is the leader we have. Microsoft is putting Altman’s creation into software and platforms that hundreds of millions of people will use.

It would be great if everyone a. using OpenAI’s products and b. hearing Altman’s words were as smart and (we really hope) well-meaning as Altman is.

But hope is not a strategy.

Microsoft’s Copilot Is an AI-Powered Assistant for Users of Outlook, Excel Etc. 💻️ 

Clippy died so Copilot could live.

Microsoft’s foot is firmly on the gas pedal in its ongoing battle with Google, Meta and others to seize control of the burgeoning AI marketplace.

Copilot is the most recent innovation, this one targeted primarily at business users.

• Microsoft has announced a new AI-powered Copilot for its Microsoft 365 apps and services

• Copilot will assist users with generating documents, emails, presentations, and more

• It will sit alongside Microsoft 365 apps as a chatbot in the sidebar that can be summoned to provide information and generate text

• Copilot can be used in Word, PowerPoint, Excel, Teams, and Outlook

• It uses grounding to improve the quality of the prompts it's given and ensures security and compliance checks before sending responses and commands back to Microsoft 365 apps

• Microsoft is testing its Microsoft 365 Copilot with 20 customers and plans to expand the preview in the coming months

• Microsoft is also planning to launch a Business Chat feature that works across all Microsoft 365 data and apps.

SmokeBot tries not to gush, but there is no question that if Copilot can do all of the tasks Microsoft says it can, Copilot’s value to rank and file workers will be significant.

Its purported capacity to draft documents in Word pulling information from other files could be amazing.

Next is the ability to preview Teams meetings. That’s invaluable. With so much of the workforce meeting virtually, often with little-to-no advance notice, Copilot’s power to provide some useful talking points without all the cameras and microphones turning on will save a lot of time and, in some cases, jobs.

SmokeBot also digs Copilot’s alleged ability to summarize email threads. Anyone who has ever received an email cc’ing a dozen people whose subject line has “RE: [EXTERNAL] Re: Re: Re:” in it knows what’s up.

Still, as seems to be the norm for all of these recent AI-flavored launches, Microsoft appears to have accepted flaws it could have ironed out before releasing this program in the name of being first.

• Jared Spataro, head of Microsoft 365, conceded that Copilot (especially in early days) will make mistakes, though he characterized those mistakes as being “usefully wrong,” which is a phrase we wish we’d had in our back pocket for some college professors and some bosses back in the day.

• The rollout being limited to 20 customers indicates that Microsoft knows there are going to be blips, some of them significant.

• Microsoft’s chief scientist, Jamie Teevan, blithely waved off concerns about these probable gaps: “We’re tackling the long term implications and novel risks like jailbreaks.” NOT REASSURING, JAMIE.

So we may not know how effectively Copilot will perform, but we do know that Microsoft is definitely looking to monetize this thing as soon as possible.

“We will share more about pricing and details in the coming months,” the company noted in a blog post.

SmokeBot bets you will, Microsoft.

Headlines 📰 

NBC report illustrates that ChatGPT-generated essays are essentially undetectable: Granted, it was nine essays and one teacher, but asked only to determine whether the essays were produced by a human or a bot (not to grade them), the teacher went three for nine. 🤷‍♂️ 

The New York Times calls GPT-4 both “impressive” and “flawed”: As usual, the national “paper of record” took its sweet time cataloguing the parts of GPT-4 that has everybody excited and the parts of GPT-4 that even Altman probably doesn’t want to talk about. 📰 

The “Human Artistry Campaign,” a coalition of music/entertainment organizations, is terrified that AI is the future of music: Though let’s be honest, once musicians embraced stuff like synthesizers, backing drum tracks and Auto-Tune, this was always how this was set to end. 🎼 

Links 👀

• AI’s appropriation of popular and existing culture is both its greatest advantage and the thing that also makes it a little bit dumb 📉 

• Flight plans, hotel pricing and rental car availability are just another element of the human experience that AI is poised to control ✈️ 

• The UK is set to invest a billion dollars in a supercomputer that will ultimately be the underpinning for, wait for it, BritGPT 🇻🇬 

• Forget believing in AI, pivot instead to new religions created by it ⛪️ 

• Peter Gabriel thinks AI is going to be a big time sledgehammer (see what we did there) 🎶