jreynar

Agreed and zero financial payoff doesn't mean no benefit. It may be that the quality of work improves, as we saw in the transition from the mimeograph to photocopiers and ultimately laser printers powered by desktop publishing software

I completely agree. It's hard to imagine this ever being perfect both because of AI and because people make mistakes in communications all the time. In our own usage of it we've found it essential to check things that seem suspicious once they're in the company brain. But, the more important gate, and one we expect may remain longer than people would like, is that we ask people to approve changes to a company brain that are proposed. Even this isn't perfect but it is a key defense from the brain turning into slop. And it's different than this demo, which was much closer to fully automated.

Cool idea. If a journalist wants to try this on an email corpus they have (legitimate) access to, send them my way. I'd be happy to help.

Hmm...maybe I've tripped people up by calling it a wiki. You can dig into a category by following the nav links at the top, e.g. to organizations (https://www.thisandthat.chat/enron/class-organization.html) and then from there dig into a specific one, such as the California Public Utilities commission by clicking on https://www.thisandthat.chat/enron/california-public-utiliti...

Thanks, really appreciate you bringing it back from the dead and hope it provides some helpful context that didn't fit naturally into the web site.

Ha, I'm a human and we made this for other humans. But I won't complain if we get some AI readers too.

The post tracks two rather different things. AI is almost certainly the hottest topic in tech right now, just like mobile was hot after the iPhone launched or cloud was a decade ago. AI-generated text is a separate complaint, and it's the only one I think is controversial. Slop, no thanks. A human using AI tools to write better is fine by me, even if a tell or two slips through.

I already answered this, but responding led me down a path to a better answer both for me, and our product. I said last week that we were working (at this+that) on building and maintaining a db of operating knowledge for companies. A crisper answer is that we're solving the problem of "unrecorded knowledge." Apparently 42% of things known by a company are only known by one employee and three of our key features are designed to help with this but I hadn't connected the dots. 1) task extraction 2) automatic contact updating and 3) the one I wrote about before: maintaining a company's operating knowledge or "brain"

AI may indeed lead to a barbell shaped distribution - lots of people with modest skills and the need for lots of awesome people. But I don't think that's necessarily a bad thing.

One of the problems with the "middle class" in the past was that there were lots of mid-level jobs in big companies but people hit a ceiling and couldn't progress. Maybe the middle hollowing out will lead to more jobs with significant opportunity and responsibility rather than ones where you get trapped because you weren't at the company early enough or all the top jobs are filled.

I'm working on making unstructured operating knowledge maintain itself. If a customer's phone number changes, that goes in the CRM. But if a salesperson learns the customer is planning to open new locations, everyone else who talks to that customer should know, and today they mostly don't. In fact, there's probably not even a good place for that valuable nugget to be stored.

The information exists. It's in email and Slack, in threads with the customer, internal discussions, meeting transcripts. But nobody updates the customer page since the payoff is deferred. We're trying to update it automatically.

Maintenance looks doable. Cold start is the part I don't have a good answer for yet: how do you work out that a company cares about customers, competitors and products as entity types, and build those lists for them, without handing them a setup task? That's what we're tackling next.

https://www.thisandthat.chat

Reviewer burden is the right diagnosis and the rule doesn't really address it.

Oracle's own FAQ concedes that reliably telling AI content from human content is impossible. So reducing volume depends on people honestly checking the box or more likely not submitting or reviewers spotting AI tells. Some of their example tells don't seem AI-specific, just bad code smells that reviewers probably rejected for prior to gen AI.

I'm reminded of academic peer review, which has run into a similar volume problem. Since there's more writing involved, there's also a lot of slop submitted and the reviews themselves can degenerate to slop, too. It's an unsolved problem. The things that seem to be making a difference are process changes like caps on submissions per author, immediate rejection prior to review, making authors review in return, etc.