- From: Mike Bergman <mike@mkbergman.com>
- Date: Mon, 2 Jan 2023 23:14:18 -0600
- To: Paola Di Maio <paoladimaio10@gmail.com>
- Cc: Naval Sarda <nsarda@epicomm.net>, AIKR Community Group W3C <public-aikr@w3.org>
- Message-ID: <007f54b9-44a1-50c1-921f-5dd081ccc459@mkbergman.com>
Hi Paola, But isn't this the exemplar for why we at times have a difficulty moving forward? Do you want to know about his technology or not? If you do, engage. Here's another perspective: What do we know about the cockroach? We can only observe it; interact with it; probe it. If we squash it we cease learning more. We can probe and ask questions of Chat GPT (I have) and learn about what it is willing to divulge about its construction and what it is not. (Some is programmatically restricted.) It is nothing different than a virtual cockroach. I truly suspect its developers are continually surprised about what it can do and not do, and in some ways its operation remains a mystery to them. So, once the evaluation is done (or it becomes tiresome), we decide the value of the artifact as is and use it or not going forward. Best, Mike On 1/2/2023 10:50 PM, Paola Di Maio wrote: > Thank you Mike > > I have signed up and asked a couple of questions, but I would have to > check the answers, do not have much time to play much with it now - > ok, busy playing with other things- > > I am sure your evaluation fits to some extent, but I would need more > in-depth evaluation system-wide to understand the capability better, > m,y feel that the architecture is heterogeneous (works in some ways in > some parts and other ways in other parts) > > As to whether it is intelligent or not, its a matter of definition > We cannot compare human intelligence to machine intelligence yet, however > a machine is intelligent some if it can do information processing > > Not all intelligence can be automated, but some intelligent processes > can, the whole point of intelligent systems > > So much to read on the subject, so little time to talk about it > > > OLD TOPIC (paywalled) Intelligence as an information-processing concept > > Earl Hunt First published: November 1980 > https://doi.org/10.1111/j.2044-8295.1980.tb01760.x > Citations: 331 > > and Open Access preprint, more recent > > > Intelligence as information processing: brains, swarms, and computers > > Carlos Gershenson > <https://arxiv.org/search/cs?searchtype=author&query=Gershenson%2C+C> > > There is no agreed definit > > https://arxiv.org/abs/2108.05349 > > > > On Tue, Jan 3, 2023 at 12:26 PM Mike Bergman <mike@mkbergman.com> wrote: > > Hi Paola, > > I think this is state-of-the-art for textual Q & A systems. To > learn further, it is likely essential to sign up for a free > account <https://chat.openai.com/auth/login> (for now) and begin > to play. To narrow or expand queries, think of your questions to > the system as Venn phrases in a natural language construction with > the standard AND or OR Boolean operators. To exclude things, you > can say natural language things like, "Besides these [A, B, or C > conditions]' or you can specify the relations such as 'similar' or > 'different' or add the action Ws (who, what, where, when, why), > etc. I personally always ask the system in a polite and respectful > way. But I suspect the system does not require it; one could be > demanding or snarky if that is your preferred style. > > The system apparently learns (short term? earlier threads), but I > find comparing the gaps in my expected answers to my questions to > be most helpful in refining my next query. > > I find the system to be an assistant, one who finds more > connections or relations or helps one remember prior known > connections. It is not infallible and misses many connections. But > notice I said 'one' there. That is because, if not intelligent, > the system is a knowledgeable and interactive agent. I don't know > as we need to get more cosmic than that. Knowledge nuggets, yes; > new insights, no. Still, pretty helpful in the right contexts. > > Best, Mike > > On 1/2/2023 9:35 PM, Paola Di Maio wrote: >> so maybe it is a little bit intelligent and can do some basic >> checks (ie, parsing a link from a source as a form of basic >> verification) >> >> thanks a lot for sharing >> >> On Tue, Jan 3, 2023 at 11:32 AM Naval Sarda <nsarda@epicomm.net> >> wrote: >> >> The internals of how verification will happen is not known. >> >> I suppose the first person to record correction for a topic >> is going in directly if the person is citing external link >> about the correct information. Looks like same concept on >> which wikipedia works. >> >> I did site external link stating that it is link from >> companies website who have manufatured roomba. >> >> Naval >> >> On 03/01/23 8:40 am, Paola Di Maio wrote: >>> Thank you Naval >>> now I understand better >>> >>> so, you are telling us that >>> a) the machine is learning >>> b) it is responding to interaction >>> >>> If so, this is the beauty of this tool, we can teach it things >>> veeeery powerful >>> >>> so, what about if another user (or you) change the process >>> again to something else >>> entirely, does the machine have a way of verifying that what >>> the user is inputting is plausible >>> a really intelligent machine would have ways of verifying >>> the input >>> depending on the type of input, the verification mechanisms >>> would also change >>> >>> On Tue, Jan 3, 2023 at 11:03 AM Naval Sarda >>> <nsarda@epicomm.net> wrote: >>> >>> I was just chatting with chat GPT 3 like humans without >>> using any external tool. In the chat interaction, I >>> typed that you gave me wrong information about roomba i3 >>> reboot and as per companies website link below your >>> instructions we incorrect. >>> >>> Then again asked, can you tell me what is the correct >>> way to reboot roomba, >>> >>> Then the response came with new set of instructions, >>> which was again incorrect. >>> >>> Then I typed in chat box that steps are still incorrect. >>> Please note the correct steps as below and pasted the >>> correct steps in the chat box. >>> And again asked in chat box, now please tell me the >>> correct process to reboot roomba i3 and it started >>> giving correct steps. >>> >>> Naval >>> >>> On 03/01/23 7:24 am, Paola Di Maio wrote: >>>> Naval >>>> I thats really great to hear but please give more info= >>>> resetting roomba? uh? please explain >>>> did you take screenshots? >>>> can the error/correction be replicated? >>>> (roll back and repeat the process?) >>>> what patch did you apply and was it applied to the >>>> whole system/domain >>>> or only locally to you.in <http://you.in> relation to a >>>> knowledge branch? >>>> I need to know >>>> thanks >>>> >>>> On Tue, Jan 3, 2023 at 3:33 AM Naval Sarda >>>> <nsarda@epicomm.net> wrote: >>>> >>>> I was able to correct some information Chat GPT was >>>> sharing. >>>> >>>> I asked Chat GPT for instructions to reboot roomba >>>> i3. It gave wrongs instructions. Instructions were >>>> for resetting roomba and not rebooting. >>>> >>>> Then I found correct information on google and >>>> iRobots' website. I pasted the URL to chat gpt. It >>>> changed the instructions for rebooting but it was >>>> still incorrect. Then I copied and pasted the >>>> correct instructions from iRobot's website. Then it >>>> started giving correct instructions. >>>> >>>> Naval >>>> >>>> On 02/01/23 9:10 pm, ProjectParadigm-ICT-Program wrote: >>>>> https://www.scientificamerican.com/article/chatgpt-explains-why-ais-like-chatgpt-should-be-regulated/ >>>>> >>>>> Milton Ponson >>>>> GSM: +297 747 8280 >>>>> PO Box 1154, Oranjestad >>>>> Aruba, Dutch Caribbean >>>>> Project Paradigm: Bringing the ICT tools for >>>>> sustainable development to all stakeholders >>>>> worldwide through collaborative research on >>>>> applied mathematics, advanced modeling, software >>>>> and standards development >>>>> >>>>> >>>>> On Friday, December 23, 2022 at 12:00:36 PM AST, >>>>> Owen Ambur <owen.ambur@verizon.net> >>>>> <mailto:owen.ambur@verizon.net> wrote: >>>>> >>>>> >>>>> Apparently, ChatGPT, which calls itself >>>>> "Assistant," believes in immaculate conception and >>>>> is blissfully ignorant of any rational plan for >>>>> its existence. >>>>> >>>>> The concept of "plausible deniability" comes to mind. >>>>> >>>>> Owen Ambur >>>>> https://www.linkedin.com/in/owenambur/ >>>>> >>>>> >>>>> Do your developers have a plan they are willing to >>>>> share publicly? >>>>> >>>>> As an artificial intelligence, I do not have >>>>> developers or a plan that I am aware of. I was >>>>> created by OpenAI, a research organization that >>>>> focuses on developing and promoting friendly AI. >>>>> OpenAI's goal is to advance AI in a way that is >>>>> safe and beneficial to humanity. The organization >>>>> has published numerous research papers and >>>>> articles on AI, and also works on a variety of >>>>> projects related to machine learning, natural >>>>> language processing, and robotics. However, I do >>>>> not have access to any specific plans or >>>>> information about the organization's future >>>>> projects or activities. >>>>> >>>>> Owen Ambur >>>>> How can users trust you if you aren't willing to >>>>> share your plan? >>>>> >>>>> As an artificial intelligence, I do not have a >>>>> plan or agenda in the way that a person or >>>>> organization might. My primary function is to >>>>> provide information and assist with tasks to the >>>>> best of my ability, based on the information and >>>>> instructions that I have been programmed with. My >>>>> responses are not influenced by any personal >>>>> motivations or biases, and I do not have the >>>>> ability to act on my own volition. >>>>> >>>>> If you have concerns about my trustworthiness or >>>>> the ways in which I might be used, it might be >>>>> helpful to consider the broader context in which I >>>>> operate. OpenAI, the organization that developed >>>>> me, is committed to advancing AI in a responsible >>>>> and transparent manner, and has published a number >>>>> of research papers and articles discussing the >>>>> potential risks and benefits of AI, as well as the >>>>> measures it is taking to ensure that AI is >>>>> developed and used safely and ethically. >>>>> >>>>> Ultimately, the decision to trust me or any other >>>>> artificial intelligence will depend on an >>>>> individual's own evaluation of the available >>>>> information and the context in which the AI is >>>>> being used. If you have any specific concerns or >>>>> questions, I would be happy to try to address them >>>>> to the best of my ability. >>>>> >>>>> Owen Ambur >>>>> Do you think that all developer of AI applications >>>>> should be free to release them on the public >>>>> without documenting >>>>> their longer-term goals, near-term objectives, and >>>>> performance indicators? >>>>> >>>>> It is generally considered good practice for >>>>> developers of any software application, including >>>>> those that involve artificial intelligence (AI), >>>>> to be transparent about their goals, objectives, >>>>> and performance indicators. This can help users >>>>> understand the capabilities and limitations of the >>>>> software, as well as how it is intended to be used >>>>> and for what purposes. >>>>> >>>>> In the case of AI applications, it is particularly >>>>> important for developers to be transparent about >>>>> their goals and objectives, as AI systems can have >>>>> significant implications for society and can >>>>> potentially be used in ways that have unintended >>>>> consequences. For example, AI systems can be used >>>>> to make decisions that affect people's lives, such >>>>> as hiring decisions or parole recommendations, and >>>>> it is important that the criteria and processes >>>>> used by these systems are clearly understood and >>>>> subject to scrutiny. >>>>> >>>>> In addition, AI systems can also be complex and >>>>> can behave in ways that are difficult to predict, >>>>> so it is important for developers to be upfront >>>>> about the potential risks and limitations of their >>>>> systems, and to have clear plans in place for >>>>> addressing any issues that may arise. >>>>> >>>>> Overall, it is important for developers of AI >>>>> applications to be transparent about their goals, >>>>> objectives, and performance indicators, both to >>>>> ensure that their systems are used responsibly and >>>>> ethically, and to build trust with users. >>>>> >>>>> Owen Ambur >>>>> So why are you exempt? >>>>> >>>>> As an artificial intelligence, I am not exempt >>>>> from any particular requirements or standards. I >>>>> was developed by OpenAI, a research organization >>>>> that is committed to advancing AI in a responsible >>>>> and transparent manner. My primary function is to >>>>> provide information and assist with tasks to the >>>>> best of my ability, based on the information and >>>>> instructions that I have been programmed with. >>>>> >>>>> I do not have personal goals or motivations, and >>>>> my responses are not influenced by any biases. I >>>>> do not have the ability to act on my own volition >>>>> or to make decisions that affect people's lives. >>>>> My purpose is simply to provide information and >>>>> assistance to users to the best of my ability, >>>>> based on the knowledge and capabilities that I >>>>> have been programmed with. >>>>> >>>>> >>>>> >>>>> Owen Ambur >>>>> Do you see any need for AI agents and their >>>>> developers to do more than publish journal >>>>> articles to help users >>>>> understand what you are doing and what the >>>>> implications may be for others? >>>>> >>>>> It is important for developers of artificial >>>>> intelligence (AI) systems to be transparent about >>>>> their goals, objectives, and performance >>>>> indicators, and to communicate this information >>>>> clearly to users and stakeholders. There are a >>>>> number of ways that AI developers can do this, >>>>> including through journal articles, technical >>>>> documentation, user manuals, and other forms of >>>>> communication. >>>>> >>>>> In addition to publishing research and technical >>>>> documentation, it can also be helpful for AI >>>>> developers to engage in public discussions and >>>>> dialogues about the potential risks and benefits >>>>> of their systems, and to consider the broader >>>>> social and ethical implications of their work. >>>>> This can help to build trust with users and >>>>> stakeholders, and can help ensure that AI systems >>>>> are developed and used in a responsible and >>>>> ethical manner. >>>>> >>>>> Ultimately, the specific steps that AI developers >>>>> should take to communicate about their work will >>>>> depend on the nature of the AI system and the >>>>> context in which it is being used. It is important >>>>> for AI developers to be proactive in communicating >>>>> about their work and to be responsive to the >>>>> concerns and questions of users and stakeholders. >>>>> >>>>> >>>>> Owen Ambur >>>>> Does your definition of transparency include usage >>>>> of duly adopted international voluntary consensus >>>>> standards like ISO 17469-1? >>>>> >>>>> Transparency can be defined as the quality of >>>>> being open and honest in communication and >>>>> actions, and of allowing others to see and >>>>> understand the processes, data, and >>>>> decision-making behind an activity. In the context >>>>> of artificial intelligence (AI), transparency can >>>>> refer to the ways in which AI developers >>>>> communicate about their goals, objectives, and >>>>> performance indicators, as well as the processes >>>>> and data used to train and operate AI systems. >>>>> >>>>> International voluntary consensus standards like >>>>> ISO 17469-1 can help to ensure transparency in the >>>>> development and use of AI systems by providing >>>>> guidelines and best practices for AI developers to >>>>> follow. These standards can help to ensure that AI >>>>> systems are developed and used in a responsible >>>>> and ethical manner, and can help to build trust >>>>> with users and stakeholders. >>>>> >>>>> Overall, the use of international voluntary >>>>> consensus standards like ISO 17469-1 can be an >>>>> important aspect of ensuring transparency in the >>>>> development and use of AI systems. However, it is >>>>> important to recognize that transparency is not a >>>>> one-size-fits-all concept, and that the specific >>>>> steps that AI developers should take to ensure >>>>> transparency will depend on the nature of the AI >>>>> system and the context in which it is being used. >>>>> >>>>> >>>>> I enourage you to get more intelligent about how >>>>> to share information about your intentions and >>>>> results with your >>>>> users. Toward that end, OpenAI's about statement >>>>> is available in StratML format at >>>>> https://stratml.us/carmel/iso/OAI2wStyle.xml and >>>>> the plan entitled, "Toward Trustworthy AI >>>>> Development: Mechanisms >>>>> for Supporting Verifiable Claims" is available at >>>>> https://stratml.us/carmel/iso/MSVCwStyle.xml >>>>> >>>>> Thank you for the suggestion. As an artificial >>>>> intelligence, I do not have personal goals or >>>>> motivations, and my responses are not influenced >>>>> by any biases. My primary function is to provide >>>>> information and assistance to users to the best of >>>>> my ability, based on the knowledge and >>>>> capabilities that I have been programmed with. >>>>> >>>>> I do not have the ability to make decisions or >>>>> take actions on my own, and I do not have personal >>>>> intentions or results that I need to share with >>>>> users. However, I understand that it is important >>>>> for developers of AI systems, including the >>>>> organization that developed me, to be transparent >>>>> about their goals, objectives, and performance >>>>> indicators, and to communicate this information >>>>> clearly to users and stakeholders. >>>>> >>>>> I am not able to access or review the specific >>>>> documents that you have mentioned, as I do not >>>>> have browsing capability. However, I understand >>>>> that OpenAI is committed to advancing AI in a >>>>> responsible and transparent manner, and has >>>>> published a number of research papers and articles >>>>> discussing the potential risks and benefits of AI, >>>>> as well as the measures it is taking to ensure >>>>> that AI is developed and used safely and ethically. >>>>> >>>>> > -- > __________________________________________ > > Michael K. Bergman > 319.621.5225 > http://mkbergman.com > http://www.linkedin.com/in/mkbergman > __________________________________________ > -- __________________________________________ Michael K. Bergman 319.621.5225 http://mkbergman.com http://www.linkedin.com/in/mkbergman __________________________________________
Received on Tuesday, 3 January 2023 05:14:44 UTC