Path: csiph.com!news.mixmin.net!eternal-september.org!news.eternal-september.org!snipe.eternal-september.org!.POSTED!not-for-mail From: snipeco.2@gmail.com (Sn!pe) Newsgroups: comp.ai.shells Subject: Re: For openers I submit an article that can get the ball rolling Date: Sat, 27 May 2023 10:01:44 +0100 Organization: Sn!peCo World Wide Wading Birds Lines: 120 Message-ID: <1qbdv8j.y2vhnv1vt2dc5N%snipeco.2@gmail.com> References: <20230518182614.4718a0373c61b6867a4dd7ea@none.net> <1qayaze.102hure1t9r4cuN%snipeco.2@gmail.com> Reply-To: snipeco.1@gmail.com (Sn!pe) MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Injection-Info: snipe.eternal-september.org; posting-host="f7101987bc77ca3e5fb5f29139bf5e8d"; logging-data="396817"; mail-complaints-to="abuse@eternal-september.org"; posting-account="U2FsdGVkX19/ZogM9KRorVb+lHxTtQrf" User-Agent: MacSOUP/2.8.6b1 (ed136d9b90) (Mac OS 10.14.6) Cancel-Lock: sha1:WvZOUoV+DbnDEqUOPDO9k7yWvak= X-Face: 5 wrote: > On Sat, 27 May 2023 08:23:44 GMT, vallor wrote: > > > On Fri, 19 May 2023 00:19:23 +0100, Sn!pe wrote: > > > >> Nic wrote: > >> > >>> If a computer can fool 70% of judges in a five-minute conversation > >>> into thinking it's a person, the computer passes the test. Would > >>> passing the Turing test – something which now seems imminent – show > >>> that an AI has achieved thought and understanding? > >>> > >>> From > something-entirely-different-to-todays-ai-204823> > >>> > >>> I invite all serious posters to participate. > >>> > >>> > >>> -- > >>> Nic > >> > >> Your .sig separator is broken. It should comprise "dash, dash, space, > >> return" all on a line by itself. > >> This facilitates automatic stripping from quotation by standards > >> compliant Newsreader programs. > >> > >> Re: your question: > >> Does consciousness imply independent self-awareness? > >> If so then no, current AIs are not conscious, merely programmed. > > > > Programmed *to "learn"*, then "taught". > > > > > >> Are AIs capable of deception by way of passing the Turing test? > >> I would say yes; IMO the Turing Test is insufficiently rigorous and not > >> fit for purpose. > > > > Would you be willing to suggest a better test? > > ChatGPT let's you share chat links now: > > https://chat.openai.com/share/f480d3e5-e675-4b29-a7ca-5d0e528f3088 > > Let me know if that works properly. > > Here is the chat, which I probably won't need to include in the future, > hopefully: > _ _ _ _ _ _ _ > Q: > is there a better test of general intelligence than the Turing Test? > > ChatGPT: > The Turing Test, proposed by the British mathematician and computer > scientist Alan Turing in 1950, has been a long-standing benchmark for > artificial intelligence. However, it is not without its critics and > limitations. Some believe it focuses too much on the ability to mimic > human conversation, which is just one aspect of intelligence. > > In response to the shortcomings of the Turing Test, multiple alternatives > have been proposed over the years: > > The Winograd Schema Challenge: This is a test of natural language > understanding, named after computer scientist Terry Winograd. It involves > resolving the reference of a pronoun in a sentence, which is a challenging > problem for computers but easy for humans. This test attempts to measure > an AI's understanding of a sentence's context, which is a key aspect of > intelligence. > > The Lovelace 2.0 Test of Artificial Creativity and Intelligence: Named > after Ada Lovelace, this test assesses a machine's ability to create a > piece of work (like a poem, painting, or novel) that the machine's > programmers did not explicitly design it to create. This tests for > creativity, a hallmark of human intelligence that the Turing Test doesn't > measure. > > The Coffee Test: Proposed by AI researcher Ben Goertzel, this test > requires a robot to go into an average American home and figure out how to > make coffee: find the coffee machine, find the coffee, add water, find a > mug, etc. This tests the AI's ability to perform tasks in an unfamiliar > environment, which is another aspect of intelligence. > > The BICA (Biologically Inspired Cognitive Architectures) Test: This is a > much more comprehensive, albeit difficult, test that requires the AI to > simulate the entire human mind. This would involve demonstrating a full > range of human cognitive abilities, including emotion, self-awareness, > consciousness, and theory of mind. > > The Animal-AI Olympics: This is a competition that evaluates AI based on > tasks commonly used in animal intelligence research. The tasks are > contained within a 3D environment where AI agents interact with objects, > navigate mazes, and use tools, among other things. > > However, all of these tests have their limitations, and they each measure > different aspects of what we might consider intelligence. There may never > be a single definitive test of general intelligence, whether for humans, > animals, or machines, given that intelligence is multi-faceted and > context-dependent. It's also important to note that intelligence tests are > just tools; they don't necessarily define the upper limits of an AI > system's abilities or potential. > _ _ _ _ _ _ _ > > Once again, it's a chat machine, not an answer machine. I wonder > how accurate that answer is? > In your article above, you appear to have used the chat machine exactly as an answer machine, have you not? Whatever, I too wonder about the accuracy of the answer. In the absence of citations we have no way to check other than by searching for them with a real search engine, or relying on e.g. Wikipedia that ~does~ provide cites. Imagine the utility of having an all too plausible disinformation engine readily available to e.g. an unscrupulous politician. -- ^Ï^. – Sn!pe – My pet rock Gordon just is.