Sponsor Banner

Collapse

Is Anyone Else Afraid of How AI Might Be Used?

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts
  • davewerden
    Administrator
    • Nov 2005
    • 11142

    Is Anyone Else Afraid of How AI Might Be Used?

    I'm seeing a lot of excitement about Artificial Intelligence (AI) these days. And many folks are toying with it for harmless things. One example is asking AI to create a picture of a trumpet, which usually comes out as something missing the basic input and output consideration!

    Our society and governments are perfectly capable of jumping in too soon with technology (and also sometimes more too slowly). How about if we just eliminate the huge cost of running the IRS and let AI decide how much each person should pay? (I have heard no talk suggesting this, just to be clear.)

    Anyway, I found an encouraging use of AI in the newish Perplexity site. It has done pretty well in the basic questions I gave it (difference between baritone and euphonium, for example) and has a link to sources where appropriate, including this website.

    So I took a tougher approach tonight and asked it if "euphonium performance" is a good college major for a later career. Here is the response:

    Majoring in euphonium performance can be a good college major as a career choice for some individuals. However, it is essential to consider the potential career options and challenges associated with this major. Some possible career paths for euphonium performance majors include:
    Military bands: A common career option for euphonium performance graduates is to join a military band
    1
    .
    Teaching: Many euphonium performance majors go on to teach music, either as private instructors or in educational institutions
    1
    2
    .
    Performing: Some graduates may pursue a career as a professional performer in classical music, although this path can be competitive and may require a strong network and entrepreneurial skills
    2
    .
    While pursuing a degree in euphonium performance can be rewarding, it is crucial to consider the potential challenges and limitations. The job market for euphonium performance graduates may be smaller and more competitive than other fields, and success may require a combination of talent, dedication, and luck. Additionally, the income potential for euphonium performers and teachers can vary widely, and financial stability may not be guaranteed.
    In conclusion, majoring in euphonium performance can be a good college major for those who are passionate about the instrument and willing to work hard to develop their skills. However, it is essential to be realistic about the potential career options and challenges associated with this major and to consider alternative career paths or additional education to increase employability and income potential.
    The numbers were actually superscript with links to 1 (dwerden.com) and 2 (reddit).
    Dave Werden (ASCAP)
    Euphonium Soloist, U.S. Coast Guard Band, retired
    Adams Artist (Adams E3)
    Alliance Mouthpiece DC3, Wick 4AL, Wick 4ABL
    YouTube: dwerden
    Facebook: davewerden
    Twitter: davewerden
    Instagram: davewerdeneuphonium
  • SteveMcGovern
    Member
    • Mar 2006
    • 98

    #2
    Those things always read like a high school student BS'ing his way through an essay.

    Comment

    • ghmerrill
      Senior Member
      • Dec 2011
      • 2387

      #3
      Originally posted by SteveMcGovern View Post
      Those things always read like a high school student BS'ing his way through an essay.
      And for good reason: they ARE like a high school student BS'ing his way through an essay. They use the same techniques that student would use in acquiring information ("knowledge"), arranging it, selecting pieces of it relevant to his assignment, and then editing it into a coherent presentation -- and this can include actually re-writing the original information that's been retrieved.

      This is not trivial. It is, in fact, a huge success, and not an empty accomplishment by engineers and cognitive scientists. It has taken time. When I first started working for Glaxo Wellcome in 1997 as a "knowledge engineer" I was heavily involved in efforts to make use of the Cyc system (https://cyc.com/) for use in scientific information retrieval, and was working closely with the Cyc engineers, linguists, and scientists to develop prototype AI software for use in drug discovery and bioinformatics. The founder of Cyc was Doug Lenat (recently deceased in Sept. 2023), and his vision and goal for decades had been to produce an AI capable of "common sense" reasoning at the level of a 12-year old human. We're not there quite yet, but we're getting close. The huge leaps forward in hardware (memory capacity, processor speed, etc.) have made possible what we're seeing today. Doug always said that we were being constrained by the available hardware, and when that was improved there would be huge leaps forward. There have been, in the areas of AI perception (think how self-driving cars sense and react to their environment), natural language understanding by AIs, reasoning (yes, "thinking") and robotics.

      What most people see of this is at the surface level in such things as robotics, autonomous self-driving cars, some funny/scary robots (I love the robot dogs!), and of course ChatGPT. All of these are inching their ways towards Doug's (and others') visions of true artificial intelligence.

      But we're not there yet. The applications we currently see (such as ChatGPT and Perplexity) can do knowledge acquisition, knowledge organization, extraction, and summary. They can in fact (I assume, though I'm not sure of this) make certain kinds of "discoveries" using this knowledge (that is ... I know how an AI like that can be created, and have done it in certain contexts). But they are still limited in terms of full reasoning and knowledge discovery.

      However, to get back to Dave's original question in his title ... I'm not really "afraid" of AI -- yet. I am concerned about how certain segments of our society might want to use it -- and any tool can be misused. And we really need to be careful about how far we go with achieving totally autonomous reasoning systems and then giving them power that would be difficult or impossible for us to control. And if people really knew about how powerful the systems are that we currently have, they'd be even more concerned -- and that doesn't count the ones that we don't know about but, that surely exist. I've given papers at AI conferences where some really spooky projects have been described by some folks in the defense sector. But I won't go droning on about this any longer. I'm out of that game now, and spending a lot of time playing trombone and euphonium.
      Last edited by ghmerrill; 01-07-2024, 02:11 PM.
      Gary Merrill
      Wessex EEb Bass tuba (DW 3XL or 2XL)
      Mack Brass Compensating Euph (DE N106, Euph J, J9 euph)
      Amati Oval Euph (DE 104, Euph J, J6 euph)
      1924 Buescher 3-valve Eb tuba (with std US receiver), Kelly 25
      Schiller American Heritage 7B clone bass trombone (DE LB K/K10/112/14 Lexan, Brass Ark MV50R)
      1947 Olds "Standard" trombone (Olds #3)

      Comment

      • ghmerrill
        Senior Member
        • Dec 2011
        • 2387

        #4
        If you're interested in some of the problems and issues in the most popular current approaches to AI, here's a paper by Doug Lenat and Gary Marcus (psychologist/cognitive scientist/neuroscientist) that lays out what they see as requirements for a trustworthy AI. It's not a particularly unbiased perspective, and is in fact largely a defense of the approach that Lenat and Cycorp took in developing the Cyc system -- which, we have to concede, failed to achieve the significant advances that have been made over the past 25 years in moving towards practical and powerful autonomous AI systems using techniques such as neural networks and machine learning rather than rule-based reasoning. But there are some good criticisms in it, and some suggestions for how to avoid what many see as the dangers of AI. It's readable by "normal people," but it is somewhat vague in places and a little heavy going in others. But skimming it may be of interest. It does mention problems with ChatGPT and chatbots. If you read it, think about Isaac Asimov's "I, Robot" and his "Three Laws of Robotics."
        Full disclosure: By 2001 I was moving away from trying to use Cyc as a basis for a scientific reasoning system to aid drug discovery, drug safety, and data and text mining of pharmaceutical data -- precisely because I couldn't get Doug to deviate from his somewhat quixotic goal of producing an AI capable of the common sense reasoning of a 12 year old human child. That wasn't what I needed, and when I moved to Novartis and then (back to) GlaxoSmithKline I went in a completely different direction in AI, though I still recognize the value and importance of Doug's approach -- and the importance of his thinking on some of the dangers of autonomous AI systems.
        Last edited by ghmerrill; 01-07-2024, 09:54 AM.
        Gary Merrill
        Wessex EEb Bass tuba (DW 3XL or 2XL)
        Mack Brass Compensating Euph (DE N106, Euph J, J9 euph)
        Amati Oval Euph (DE 104, Euph J, J6 euph)
        1924 Buescher 3-valve Eb tuba (with std US receiver), Kelly 25
        Schiller American Heritage 7B clone bass trombone (DE LB K/K10/112/14 Lexan, Brass Ark MV50R)
        1947 Olds "Standard" trombone (Olds #3)

        Comment

        • hollowayp
          Junior Member
          • Jun 2023
          • 2

          #5
          HereÂ’s a viewpoint from one of the music technology companies:
          https://articles.roland.com/how-musi...an-utilize-ai/

          Comment

          • ghmerrill
            Senior Member
            • Dec 2011
            • 2387

            #6
            For those who might be interested in a fairly intelligible history of AI devoid of complex technical details ...

            You might be interested in looking at this:

            AI Winter

            It's at a pretty high descriptive level, but at least doesn't contain the confusions, conflations, misunderstandings, and drivel that articles by hucksters and marketeers do. It sketches the history and development of AI (in terms of goals, successes, and failures) over about 50 years. It provides a brief view of how far things have come, but without an account of exactly how that happened. So it's a bit like the history of the automobile from Henry Ford to today, by looking at various models across the years.

            I note that at some point when I wasn't looking, the meaning of the acronym AAAI was changed from "The American Association for Artificial Intelligence" to "The Association for the Advancement of Artificial Intelligence". In part, this was because the older name of the association made it sound like a club whose members were robots. Of course, there's still the Association for Computing Machinery which really does sound like a club for robots.

            What has struck me most about it, looking back across those years at AI evolution, is how old I am. But also how interesting it was to be involved in some of those pieces: natural language understanding, expert systems, informatics, cognitive systems, knowledge-based systems, learning systems, and (not mentioned in the article) text and data mining. Unfortunately, the article stops a bit short of the status of AI today, doesn't delve into the hardware advances (memory size, processor speed, and size of components), and only briefly mentions the rise of "generative AI" and learning (or self-modifying) systems. For a similar high level view of generative AI, see

            Generative artificial intelligence
            Gary Merrill
            Wessex EEb Bass tuba (DW 3XL or 2XL)
            Mack Brass Compensating Euph (DE N106, Euph J, J9 euph)
            Amati Oval Euph (DE 104, Euph J, J6 euph)
            1924 Buescher 3-valve Eb tuba (with std US receiver), Kelly 25
            Schiller American Heritage 7B clone bass trombone (DE LB K/K10/112/14 Lexan, Brass Ark MV50R)
            1947 Olds "Standard" trombone (Olds #3)

            Comment

            • DaveBj
              Senior Member
              • Oct 2011
              • 1064

              #7
              Getting back to the title of the thread, whenever I think of AI, I think of the Rise of the Machines in the Dune Universe, and the ensuing Butlerian Jihad, whose motto was "Thou shalt not make a machine in the image of the human mind.
              David Bjornstad

              1923 Conn New Wonder 86I, Bach 6 1/2 AL
              2018 Wessex EP100 Dolce, Denis Wick 4ABL
              2013 Jinbao JBEP-1111L, Denis Wick 4AM
              2015 Jinbao JBBR-1240, Denis Wick clone mouthpiece of unknown designation
              Cullman (AL) Community Band (Euph Section Leader)
              Brass Band of Huntsville (2nd Bari)

              Comment

              • ghmerrill
                Senior Member
                • Dec 2011
                • 2387

                #8
                Originally posted by DaveBj View Post
                "Thou shalt not make a machine in the image of the human mind.
                For several reasons, I'm inclined to think that those are in fact the best and safest to make -- at least initially. The problem is that once you can make a self-modifying machine, how can you (and should you) impose limits on that? And once you arguably cross the threshold of sentience (or at least self-awareness), how should those be treated in society? These may seem like fanciful questions, but as "AI" and "machine intelligence" advances on the one hand, and as (particularly human) neuroscience advances on the other, they're not as fanciful as you might suppose. And we do currently have self-modifying machines in one or another sense (have had for some time -- a major point of the LISP programming language was to facilitate that). So the future is very near -- if not right in front of us.
                Gary Merrill
                Wessex EEb Bass tuba (DW 3XL or 2XL)
                Mack Brass Compensating Euph (DE N106, Euph J, J9 euph)
                Amati Oval Euph (DE 104, Euph J, J6 euph)
                1924 Buescher 3-valve Eb tuba (with std US receiver), Kelly 25
                Schiller American Heritage 7B clone bass trombone (DE LB K/K10/112/14 Lexan, Brass Ark MV50R)
                1947 Olds "Standard" trombone (Olds #3)

                Comment

                • ghmerrill
                  Senior Member
                  • Dec 2011
                  • 2387

                  #9
                  To see how sensitive these issues are becoming, consider the following in relation to ChatGPT:

                  ""In June 2022 the Google LaMDA (Language Model for Dialog Applications) chatbot received widespread coverage regarding claims about it having achieved sentience. Initially in an article in The Economist Google Research Fellow Blaise Agüera y Arcas said the chatbot had demonstrated a degree of understanding of social relationships.[47] Several days later, Google engineer Blake Lemoine claimed in an interview with the Washington Post that LaMDA had achieved sentience. Lemoine had been placed on leave by Google for internal assertions to this effect. Agüera y Arcas (a Google Vice President) and Jen Gennai (head of Responsible Innovation) had investigated the claims but dismissed them.[48] Lemoine's assertion was roundly rejected by other experts in the field, pointing out that a language model appearing to mimic human conversation does not indicate that any intelligence is present behind it,[49] despite seeming to pass the Turing test. Widespread discussion from proponents for and against the claim that LaMDA has reached sentience has sparked discussion across social-media platforms, to include defining the meaning of sentience as well as what it means to be human."

                  So ... an engineer at Google was "placed on leave" (basically, suspended from his job!) for suggesting that ChatGPT was sentient.

                  Now, I'll grant you that this was not a smart thing to say and that Google doesn't want to be put in the position of having to explain or defend such assertions. Also, ChatGPT is very far from being sentient or showing any genuine evidence that it is. But notice, inter alia here, it's conceded that ChatGPT did appear to pass the Turing test -- which, though not offered by Turing as a test of sentience, has traditionally been regarded as the qualifying benchmark for intelligence (which are somewhat conflated in this context). Also notice, that this was serious enough that a Google VP (and Fellow at Google Research) and Google's Director of Responsible Innovation (now head of Google's AI Ethics & Compliance) felt compelled to "investigate" the sentience claim and dismiss it as unfounded.

                  Very few -- if any -- in the AI industry are casually denying that an AI is (or may be) intelligent in at least certain clearly reasonable senses of that term, and the commitment to a view that an AI could never be sentient is beginning to find refuge in theology and a few areas of speculative metaphysics.
                  Gary Merrill
                  Wessex EEb Bass tuba (DW 3XL or 2XL)
                  Mack Brass Compensating Euph (DE N106, Euph J, J9 euph)
                  Amati Oval Euph (DE 104, Euph J, J6 euph)
                  1924 Buescher 3-valve Eb tuba (with std US receiver), Kelly 25
                  Schiller American Heritage 7B clone bass trombone (DE LB K/K10/112/14 Lexan, Brass Ark MV50R)
                  1947 Olds "Standard" trombone (Olds #3)

                  Comment

                  Working...
                  X