 |
BEYOND GPT: Trusted Computing and the Future of Explainable AI By Mark Anderson Why Read: The world is quickly waking up to the sobering reality behind running software that inherently fails, such as generative AI. While softer, language-based functions (with low cost for failures) continue to grow in the marketplace, authoritative voices in business enterprise, medicine, and academia are increasingly stepping back from this technology. In this week's issue, we'll look at what comes next, as the blush truly is off the chat "rose." This issue is a transcript on the subject from the opening of Advanced AI Day at FiRe 2023, and I'm pleased to note that we got it right. I hope it is of use to our members. - mra _______
Introduction When Springer - a top publisher of science and technology papers in the EU and worldwide - issues a new paper titled "ChatGPT is Bullshit," you can make the reliable statement that the winds have turned against OpenAI's CEO, Sam Altman, and the vacuous hype behind the firm's products. Of course, having the "safety team" leave the firm and making himself - the root cause of the need for this team - the new safety leader is more evidence of a similar nature. The backdoor gossip that one of the real reasons his board fired him earlier is even more concerning, if true: did Sam really ship the world's most dangerous (and still unfinished) product without consulting his board? Either way, the University of Glasgow seems to be on to his arrogant and goofy predictions about Watson's children: ChatGPT is bullshit | Ethics and Information Technology (springer.com) At the same time, Science Journal editors issued a piece examining the multitudinous claims by AI (neural net) companies that they were discovering new drugs (someday). A deep dive on all these claims from a recent review led them to conclude that all the claims were less than true. In what has turned out to be a busy week in the AI publishing world, a German book came out (along with its English version) titled Ignite 2034, in which numerous scientists, Nobelists, and a stray tech CEO or two look 10 years ahead, each writing about their own area of interest. So, of course, my chapter is on "The Future of AI: Pattern Discovery, Explainable AI, and Ethical AI." It is available to order here, on Amazon. We are issuing this transcript today because we're now at the inflection point predicted in my interview at FiRe last November. I think you'll find host David Brin's part of this conversation equally insightful - which is to be expected, given his "chops" in the AI consulting role he often plays in the intel community. It's time to turn the page on Chat and move into the chapter on Discovery Engines. The good news for you, our members, is simple: that next page begins just below. - mra
Trusted Computing and the Future of Explainable AI Wednesday, November 8, 2023 Los Angeles, CA
David Brin, Author & Physicist: Hello [....] My name is David Brin; I'm your local - well, one of your two local sci-fi folks here. And I'm here to interview our [FiRe] host, Mark Anderson, about some of the things that have been vexing us in public - I've been involved in these debates online also, about AI - and how something he's been working on with several of you, Pattern Computer [of which Mark is CEO], might offer ways around some of the things that have been vexing us in public. Let's dive right in, Mark. Why don't you tell us what you mean by - well, we've heard about pattern computing. What do you mean by "pattern discovery," and how does it differ from regular discovery? Mark Anderson: That word was around before we picked it up and stole it and made it our own - which we've done, I think. It's the output, to be simple: if you had the most wonderful machine, or system, for finding, looking at, discovering, patterns - and learning all about them - which, by the way, is the game ... That is the game. The output would be pattern discoveries. This was a theory that came to me from the work with SNS, and it only became clear on a compute landscape when we built the first Pattern computer. I had learned earlier from working on pattern recognition, and doing SNS work, that the wider the funnel, the more data, the more variety of types of data you use, the longer you study something, the better your memory, the better you can sort these things into boxes and so forth. You suddenly find things - like the Chinese national business model - that nobody else had. By accident. And it's like, "Woah!" That's a pattern discovery.
|
