Between trip, end-of-year initiatives, the approaching holidays, and different hysteria, I haven’t give you an article this month. So right here’s a fast listing of issues which have amazed me not too long ago.
Are we digital but?
I’m removed from the primary individual to search out NotebookLM superb, and I actually gained’t be the final. I did a easy experiment: I pointed it at two of my current posts, “Assume Higher” and “Henry Ford Does AI.” Each the abstract and steered questions NotebookLM offered have been fairly good: They went past merely commenting on the 2 items and acquired into the connection between the 2. However what blew me away was the podcast it generated: an eight-minute dialogue between two artificial individuals who sounded and engaged. (Right here’s an outline of among the strategies Google places to make use of to make it occur.) Was it 100% appropriate? No, however actually, if a human summarized my articles, I’d in all probability discover a number of issues to complain about.
Study quicker. Dig deeper. See farther.
Being Google, after the preliminary expertise, the consumer interface was greater than a bit clunky. After I wished to return to the podcast a number of days later, I needed to play “guess what to click on” manner an excessive amount of. (Trace: Would you guess that you want to click on on “Pocket book Information”? Why doesn’t the podcast participant seem by default?) However that’s actually a really minor drawback.
Fashions utilizing computer systems
Anthropic’s pc use API is now out there in beta. Beta is true—there’s clearly lots happening right here that’s harmful and simply abusable. However it’s additionally a whole lot of enjoyable, and it factors towards a brand new path for AI growth.
In essence (and I could have the essence flawed), pc use lets you inform Claude tips on how to use a pc: browsers, editors, shells, something that may have a consumer interface on a display (and probably extra). Anthropic supplies a demo as a Docker container, so you possibly can run it safely. As soon as the container is operating, you may give Claude an issue to resolve; it can work out tips on how to clear up that drawback, and use the container’s digital Linux pc to do the work. For instance, you possibly can ask it to fill out a spreadsheet with knowledge it collects from web sites. Claude will do all the click, copying, and pasting for you.
Is that this revolutionary? My first response was “Massive deal, I can add a file to GPT and use it to browse the net for me.” In precept that’s true, though ChatGPT doesn’t permit net shopping and file importing in the identical dialog. What’s actually new? Take into consideration the monstrous immediate you’d must get GPT to learn a spreadsheet, discover out what knowledge was lacking, search for that knowledge on the internet, and generate a brand new up to date spreadsheet. It wouldn’t be easy. With pc use, most of that complexity disappears.
Does it actually disappear? We’ll discover out as we get additional in. We’re nonetheless on the stage the place hallucinations and misbehavior are cute fairly than important. It’s simple for Claude to be misled into decoding one thing on a random web site as a immediate. It is going to be a subject day for immediate injection assaults. And I can think about loads of enhancements. Pc use at the moment works by taking screenshots and delivery them to Claude, which computes the place to click on. That appears extremely awkward, particularly provided that many purposes have accessibility affordances which may make the screenshotting pointless.
For now, chill out and take a breath. Don’t use pc use for something severe but—it’s necessary to heed Anthropic’s many warnings. However it is best to play with it and take into consideration what it means. An automatic framework for testing net purposes, Selenium++? A instrument for negotiating with on-line distributors? We’re a lot nearer to an agent-filled world the place we ask a pc what to do and it does it for us.
Might this be the tip of CRM?
Considerably alongside the identical traces: Sam Lessin posted on Twitter (I gained’t name it X) a few very intelligent and helpful hack. He exported a few years of electronic mail, used GPT to extract key components, and uploaded it to NotebookLM (sure, once more), which permits him to ask questions on his conversations over the previous decade. Who did I discuss to? Why? What are the subjects we talked about? That’s all helpful data.
Sam argues that that is the tip of structured buyer relationship administration (CRM) software program. I gained’t supply an opinion for buyers or founders, however his course of resonated with me instantly. I’ve labored with many authors and potential authors over the a long time, and my electronic mail contains conversations with hundreds of individuals. So once I wish to ask a query like “I wish to perceive extra about DDOS; who ought to I discuss to?” my first step is to go to Gmail and begin looking. E-mail is my CRM system; I’ve by no means used a business CRM product.
Sadly and satirically, Gmail’s means to go looking is sort of poor. Utilizing it for contact administration, although it may be made to work, isn’t nice. Can I simply ask NotebookLM? Completely.
E-mail-based CRM would possibly even be a very good startup concept, although it’s laborious to think about succeeding long-term. There wouldn’t be a lot of a “moat” to guard a startup towards bigger firms—like Google itself. I can simply think about Google constructing this type of AI-enabled search immediately into Gmail. They have already got all the information.
That’s it for this month. That wasn’t so unhealthy—possibly I ought to do that extra typically.
