top of page

Why RTFM When You Can Ask chatGPT?

Since the new trick pony showed up, I've been kicking the digital tires, and like many, I'm impressed. Not so much by the "AI" part, as by the sheer utility of it. Let's consider what's wrong with the web X.0 as it exists today and why chatGPT is so much better.

If I want an answer to a technical question (bearing in mind the limitations of the platform) such as "how do i use Microsoft Power Automate to automatically create tasks in outlook?", chatGPT produces an answer instead of a list of questionable website links with content of uncertain value. That answer may not be correct, but at least there is no noise to wade through, and I can refine it to improve the results.

And it is positively amazing when it comes to assembling lists of data layered underneath common search results, such as "who are the top 10 candle makers in America by volume?" That is the sort of search that would normally take me 15 minutes to get a list and verify the contents, and in doing so, I'd have touched not only the search engine, but at least 3 websites and left a trail of web breadcrumbs resulting in candle advertisements for the next month on my social media.

Consequently, I'm using it for lots of questions these days, especially getting me to the best information sources for deeper questions. Is it right? I don't know, but it's good enough for me and it's better than what I get coming from a common search engine. Time is money and chatGPT is saving me time and getting me to the finish line faster.

Consequently, we're going to need a couple things before we can feel confident about using it for non-trivial work:

  1. We need to be able to rely on the quality of the responses. If I have to validate a response I get, or spend time correcting it, that counts against the utility.

  2. We need to be able to rely on the uncensored response. Right now, we know that certain content is filtered, which is fine, but at some point, it will need classification layers and IAM for restricted content. Some of us do, on occasion, need to know about certain chemicals that we don't want in the hands of children.

  3. We need to have the utmost confidence in those classification layers and content restriction. If one country or company is hiding results, we want to know. A V&V process and certifications will be needed, but as always, who watches the watchers? Certainly not a large accounting firm, I hope. What if we used an AI to evaluate other AI's? That's interesting. We could call it VerifEye, if that wasn't taken. Definitely not calling it Skynet.

We're also going to need to reassess how we've thinly monetized the web.

  1. We will need to pay for this wizardry. Subscription levels for AI access is fine, especially to premium content, much like we have with streaming movie services. If you want the Harvard rare books collection included in your generation, or foreign language translations, you pay extra for that privilege. Or do you? It could get very interesting with public domain material and Moore's Law. Eventually it should be free for anything other than current content.

  2. Thin monetization is gone. Dinosaur. Sorry, little Carl Clickbait, but no more easy money for you. I'm not going to waste time clicking through a dozen revenue generating websites to get to where I need to be and I definitely do not want to see adverts or be advertised to based on what I look up. So revenue for content is going to be tied directly to that content, which also means we need better controls over intellectual property and site fencing. Accessing a blog, for example, might be free, but if you want to premium internal content in your AI data set response, it will require a premium subscription. This seems fair.

The de-monetization of the thin web both concerns and delights me. Most of the noise on the web has been created precisely because of the monetization to enrich people who contribute nothing to the substance. We never needed a dozen sites all pointing at the same information just so a dozen people can get the clicks. And those people are going to need to find another way to make that money as the streams dry up.

In the meantime, I'm getting a LOT of my answers faster without click-bait, and I look forward to the next iteration with deeper knowledge and faster responses.

8 views0 comments

Recent Posts

See All

Insider security breaches occur far less often than either accidental disclosures or malicious outside activity (hacking), but they can pack a much bigger punch. Insiders are already within the first

We're all familiar with the saying, "Never judge a book by its cover," and this holds up in information security as well as anyplace. Oddly enough, in information security, so does the inverse concep

Considering a merger? Here's a starting checklist for the cybersecurity components. An appropriate level of review now prevents surprises later. This is in addition to any IT considerations. Checklist

Post: Blog2_Post
bottom of page