AI’s Big Gift to Society Is … Pithy Summaries?

Change Every Page is a term that perfectly sums up Robert Caro’s nonfiction method. The word is so related to Caro that it appears in the title of the most recent film about him and of an display of his files at the New York Historical Society. No matter how mind-numbing or difficult, it is absolutely necessary to read every line of any file relating to his theme. He has discovered that something that seems minor may reveal a whole new side to an event, provide a way to an unknown cause, or disprove a mystery about who caused a crisis or accomplishment. Over his career he has pored over literally millions of pages of documents: reports, transcripts, articles, legal briefs, letters ( 45 million in the LBJ Presidential Library alone! ). Some seemed dangerous boring, repetitive, or useless. No matter—he’d push through, paying full attention. Caro’s continuous page-turning has made his job memorable.

In the age of AI, yet, there’s a fresh motto: There’s no need to move sites at all! Not even the records of your conversations. Oh, and you do n’t have to pay attention at meetings, or even attend them. You do n’t even need to read your mail or the memos from your coworkers. In no time, you’ll have a summary to check if you feed the raw materials into a huge language model. With OpenAI’s ChatGPT, Google’s Gemini, and Anthropic’s Claude as our wingmen, description checking is what then qualifies as planning.

LLMs like to describe, or at least that’s what their authors set them about doing. Google then “auto-summarises” your documents to allow you to “quickly interpret the information that matters and promote where you focus.” AI can also summarize Google Chat conversations that are not read! With Microsoft Copilot, if you so many as hover your mouse over an Excel spreadsheet, PDF, Word dentist, or PowerPoint presentation, you’ll find it boiled over. Yes, even the compressed bullet points on a slide deck can be reduced to the more crucial stuff, right? Additionally, Meta today provides a list of all the comments made on popular articles. Zoom generates a bribe sheet and a meeting summary in real time. Summaries are then front and center on sequencing services like Otter, and the transcription itself is now in another page.

Why the encounter of summarizing? Summaries are one of the most easy and immediately useful features available at a time when we are just starting to figure out how to extract value from LLMs. Of training, they may have issues or overlook critical points. Noted. The more significant danger is that we will become stupider if we rely too heavily on reports.

Summaries, after all, are hazy charts and not the place itself. I’m reminded of the Woody Allen joke where he zipped through War and Peace in 20 hours and concluded,” It’s about Russia”. I’m not arguing that AI reports are that obscure. In fact, the reason they’re unsafe is that they’re good much. They permit you to pretend it and move on with some understanding of the subject. Not particularly strong, though.

As an example, let’s get AI-generated reports of speech audio, like what Otter does. As a journalist, I know that you lose something when you do n’t do your own transcriptions. It’s extremely time-consuming. However, you are able to identify the words and phrases used by your area. You nearly always discover something missing. You might be able to recover some of that from a very thorough checking of a text. Having all summarized, though, tempts you to look at just the sections of fast interest—at the cost of discovering treasures buried in the language.

Powerful leaders have been aware of the danger of these keys from the beginning. That’s why Jeff Bezos, when he was CEO of Amazon, banned PowerPoint from his sessions. He reportedly demanded that his associates write a thorough letter known as a “6-pager.” Writing the 6-pager required managers to think hard about what they were proposing, with every syllable essential to executing, or abandoning, their ball. All turns all six pages of the file in silence during the opening session of a Bezos meeting. No describing allowed!

To be honest, I can satisfy a rebuttal to my pain with summaries. With no work anymore, an LLM does read every website. An LLM is rapidly find the most mysterious information if you want to go beyond the description and you give it the right causes. Perhaps one day these concepts will be able to discover and uncover those stones, specifically made for your needs. If that happens, nevertheless, we’d be even more reliant on them, and our own abilities may atrophy.

Long-term, description mania might lead to an degradation of writing itself. Why bother taking the time to research the details that make persuasive reading, or create the narrative to present your wit, if you know that no one will be reading the actual wording of your emails, documents, or reports? You could also hire AI to do the writing for you, which will not mind if you request that it produce 100-page accounts. Nobody did voice their opinions because they’ll get condensing the document into a number of bullet points. If everything goes as planned, a civilization’s social output will be of the same caliber as a third-generation Xerox.

Robert Caro’s second volume in his amazing LBJ saga is years past the deadline. The cycle probably would have been much finished if LLMs had been present when he had started telling the president’s story nearly 50 years ago and he had actually used them and never turned so many pages. But no nearly as great.

I spoke with Otter’s CEO Sam Liang earlier this year. When specializing in right transcription, the company now offers a range of meeting-based Artificial tools, including of course summarization—but even rougher features, like AI avatars that can enter your meetings and work the discussion. I was wondering if this might undermine the purpose of meetings in my Plain View essay on the subject.

I’m curious to know whether the presence of AI in meetings might make people less likely to show up. Knowing that a summary will be made available may discourage people from actually showing up. Liang himself claims that only a small percentage of the meetings he is invited to take part in are held. ” As CEO of a startup, I get tons of invitations to go to meetings —oftentimes I’m double booked or triple booked”, he says. ” With Otter, I can look at my invitations and rank them. I classify them based on the content, the urgency, importance, and whether my presence adds any value or not”. Since he’s the CEO, he may find it easier to opt out. On the other hand, the boss’s presence in a meeting increases its value for those who are looking for a quick yes or no to a proposal.

Of course, the premise behind meetings is that every person’s presence adds potential value. If everyone turns to the one person who can contribute to a problem at once only finds an empty seat, defeats the purpose. However, Liang also has an AI solution for that. We’re developing an Otter Avatar system that will teach a personal model to each employee for meetings where they are sick or on vacation. We will train the avatar using your historical data, or your past meetings, or your Slack messages. The avatar can respond to that employee’s question on their behalf.

I make note of the possibility that this could spark an AI arms race. ” I’m going to send my avatar to every meeting, and so will everyone else”, I explain. Meetings will be a collection of AI avatars conversing with one another; after that, people will review the summary to see what the AIs have said to each other.

” That can happen”, says Liang. ” Of course, there are always situations where you want a personal relationship directly”.

Judith asks,” Many STEM-related autobiographies contain childhood memories of hands-on experience with transistors, radios, rocket and chemistry sets. Are kids nowadays missing out on these experiences?

Thanks for the question, Judith. You’re probably correct to say that those kid-focused science projects may have come a long way. One reason is that we’re a lot more safety ( and lawsuit ) conscious now. Some toy chemistry kits included radioactive uranium ore and poisonous sodium cyanide. Caution dictated less dangerous—and less interesting—kits.

But, as we both likely know, it is n’t safety concerns that make kids less likely to venture into hands-on experimentation. With universal access to computers, childhood scientific explorations can easily be done in simulation. I still believe that young people who are consuming knowledge of how the world works can easily get their squirmy little hands on rocketry tools, telescopes, and even fascinating chemicals. Future Richard Feynmans will search for gadgetry on the internet, at the local junkyard, or through a few clubs or knowledgeable science teachers. For those not so naturally driven, the presence of world-building tools like Minecraft and games like Civilization—as well as coding itself—provides a great gateway into the world of science. Geeks are gonna geek.

You can submit questions to [email protected]. Include ASK LEVY in the subject line.

End Times Chronicle

Old: North Carolina’s mountains were protected from the storms that afflict the state’s eastern half. The new Hurricane Alley is located in Western North Carolina.

Last but Not Least

The AI-related mess that is happening in the world’s elections is shown in this scorecard.

Bobbi Althoff mommy-blogged her way to fame—and, as she’d hoped, fortune. She explains all in our Big Interview.

As fewer people sign up for DNA testing, 23andMe’s valuation is approaching spit.

License plate readers are sifting up the information on bumper stickers, which is a ton of privacy averse.

Do n’t miss future subscriber-only editions of this column. Subscribe to WIRED ( 50 % off for Plaintext readers ) today.