Path: ...!news.mixmin.net!eternal-september.org!feeder3.eternal-september.org!news.eternal-september.org!.POSTED!not-for-mail From: Don Y Newsgroups: sci.electronics.design Subject: Re: smart people doing stupid things Date: Sun, 19 May 2024 22:02:56 -0700 Organization: A noiseless patient Spider Lines: 67 Message-ID: References: MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8; format=flowed Content-Transfer-Encoding: 7bit Injection-Date: Mon, 20 May 2024 07:02:59 +0200 (CEST) Injection-Info: dont-email.me; posting-host="5d80f1dc6dc80e940061abf9488d5484"; logging-data="4029313"; mail-complaints-to="abuse@eternal-september.org"; posting-account="U2FsdGVkX19dIm4cqVf7A4MulFNXa9NW" User-Agent: Mozilla/5.0 (Windows NT 6.1; Win64; x64; rv:102.0) Gecko/20100101 Thunderbird/102.2.2 Cancel-Lock: sha1:3ROiUVc/tHHkySLE7i3xLuip85U= In-Reply-To: Content-Language: en-US Bytes: 4743 On 5/19/2024 8:51 PM, Edward Rawde wrote: > It is my view that you don't need to know how a brain works to be able to > make a brain. That's a fallacy. We can't make a *plant* let alone a brain. > You just need something which has sufficient complexity which learns to > become what you want it to become. So, you don't know what a brain is. And, you don't know how it learns. Yet, magically expect it to do so? > You seem to think that humans have something which AI can never have. I designed a resource allocation mechanism to allow competing agents to "bid" for the resources that they needed to achieve their individual goals. The thought was that they could each reach some sort of homeostatic equilibrium at which point the available resources would be fairly apportioned to achieve whatever *could* be achieved with the available system resources (because resources available can change and demands placed on them could change as well). My thinking was that I could endow each "task" with different amounts of "cash" to suggest their relative levels of importance. They could then interactively "bid" with each other for resources; "How much is it WORTH to you to meet your goals?" This was a colossal failure. Because bidding STRATEGY is difficult to codify in a manner that can learn and meet its own goals. Some tasks would "shoot their wad" and still not be guaranteed to "purchase" the resources they needed IN THE FACE OF OTHER COMPETITORS. Others would spread themselves too thin and find themselves losing out to more modest "bidders". A human faces similar situation when going to an auction with a fixed amount of cash. If you find an item of interest, you have to make some judgement call as to how much of your available budget to risk on that item, knowing that if you WIN the bid, your reserves for other items (whose competitors are yet to be seen) will be reduced. And, if you allow this to be a fluid/interactive process where bidders can ADJUST their bids, dynamically (up or down), then the system oscillates until some bidder "goes all in". The failure is not in the concept but, rather, the implementation. *I* couldn't figure out how to *teach* (code) a strategy that COULD win as often as it SHOULD win. Because I hoped for more than the results available with more trivial approaches. AI practitioners don't know how to teach issues unrelated to "chaining facts in a knowledge base" or "looking for patterns in data". These are relatively simple undertakings that just rely on resources. E.g., a *child* can understand how an inference engine works: Knowledge base: Children get parties on their birthday. You are a child. Today is your birthday. Conclusion: You will have a party today! So, AIs will be intelligent but lack many (all?) of the other HUMAN characteristics that we tend to associate with intelligence (creativity, imagination, originality, intuition, etc.)