- From: Paola Di Maio <paoladimaio10@gmail.com>
- Date: Sun, 1 Jun 2025 08:44:12 +0800
- To: Kevin Spellman <kevinfrsa@icloud.com>
- Cc: Ambur Owen <owen.ambur@verizon.net>, W3C AIKR CG <public-aikr@w3.org>
- Message-ID: <CAMXe=SrN6h+V0cz2u04qCoh0nJv1KzkA4Ljgu580v201wuNjXw@mail.gmail.com>
What to say, Owen and Kevin Yesterday a famous phrase in the Hinton Turing Award speech kept rolling into my head 'the nail in the coffin for symbolic knowledge representation' heralding an age of non logic based machine learning I attach the url and transcript for those who may want to listen to it again, it was such a good lecture btw, Too bad it played down symbolic/logic AI Time to go back to those talks, This is where the troubles started (if not earlier) and kind of feels like it was a long time ago but 2018 ts just yesterday really I am enjoying every bit of AI, and I am also startled by its limitations (abandon logic and see what you get) Mind out, poor and deficient reasoning is not just a prerogative of AI, Humans excel and make errors. flawed conclusion and fallibility in general It is when AI becomes pervasive and starts interfering with our systems deleting our emails, rewriting our browser history that is going to be scary, when innocent people use the LLM to learn and write about a topic and do not realise that what they hear is only part of the story, however well written up and fast Again this is also true of all knowledge sources, bias is not something new, it has been part of records in world history But AI is now part of the interface that filters reality, and that is why it can become scary I have also seen bias and poor reasoning in the initiatives aimed at mitigating AI risks As long as we are aware I guess,to maintain that level of awareness in a dynamic requires paying a lot of attention to what is going on and that can only be done by a well tuned human brain https://www.annualreviews.org/content/journals/10.1146/annurev-neuro-062111-150525 I want to also make a note of the transcript of the LLM output I made a mistake in my prompt, that tried to retrieve the Turing Award lecture mentioned above and wrote 2019, and the LLM hang on to the mistake throughout its response instead of correcting it. I attach two transcripts for reference only On Sun, Jun 1, 2025 at 1:35 AM Kevin Spellman <kevinfrsa@icloud.com> wrote: > Universal AI and LLM design as a regulated government responsibility would > bring accountability, uniformity, standards and ethics. Social media and > the algorithms that violate our digital rights only come to light when we > stumble on to it. LLM’s are based on our data and we did not clearly agree > to this (or at least I didn’t). There is an opacity on how they work, how > and what they are connected to and more so the steps in place to mitigate > bias as an example. In a field that is growing in complexity and revenue, > there are fewer safeguards and people to support and enforce a standard for > public and private AI handling our data. > > Please pardon the brevity > Sent from my iPhone > > *Dr. Kevin J Spellman, FRSA, CMRS* > > On 31 May 2025, at 16:17, Owen Ambur <owen.ambur@verizon.net> wrote: > > > Paola, while it might be taken as self-serving flattery or, at least, > knowing your customer, ChatGPT's conclusion about the second of your two > references makes sense to me: > > Bottom Line > > Steven J. Vaughan-Nichols is voicing a legitimate warning: *if we train > AIs on trash, they will produce trash.* But the current reality is not > that AI is collapsing—it’s that the ecosystem around it is fragile and > poorly governed. The way forward isn't to abandon AI but to become more *intentional > and structured* in how we curate knowledge, govern inputs, and manage > usage. > > That’s where standards like StratML, structured data, and truly > responsible AI design can help avert the kind of collapse the article warns > about. > > The details of its argument are available here > <https://chatgpt.com/share/683b1bb1-14c0-800b-9d9a-381ce0935ec8>. > > Owen Ambur > https://www.linkedin.com/in/owenambur/ > > > On Saturday, May 31, 2025 at 12:10:11 AM EDT, Paola Di Maio < > paola.dimaio@gmail.com> wrote: > > > Good day > > I hope everyone gets a change to smell the flowers at least once a day > > As predicted, we are rapidly rolling into a new age of AI driven > everything and knowledge is all we ve got to understand what is happening > and how > > The changes are already impacting our individual and collective lives and > behaviours etc > and we won't even know (scratching head) > > The best that we can do is hang onto our instruments of discernment, KR > being one of them > > Two articles below bring up important points > > *Gemini may summarize your emails even if you dont opt it for the feature* > > https://techcrunch.com/2025/05/30/gemini-will-now-automatically-summarize-your-long-emails-unless-you-opt-out/ > > Honestly I do not know if this is true. It may even be illegal and if it > depends on the geographi loation could end up being very confusing > for those who travel around a lot. How will it work, if one day a person > reads an email from one country and another day from another? > if someone is a Google insider enough, should be investigated imho > > *AI Model Collapse* > https://www.theregister.com/2025/05/27/opinion_column_ai_model_collapse/ > When the AI models collapse all we are going to have left is going to be > the robust knowledge structure in our brain/minds and in our libraries > > > *Brace, brace* > > >
Attachments
- text/plain attachment: Hinton_Turing_Lecture_2018_Deep_Learning.txt
- text/plain attachment: Yan_Le_Cun_Turing_Award_Lecture_201.txt
- application/pdf attachment: Did_Hinton_and_LeCun_s_2019_Turing_lecture_challen.pdf
- application/pdf attachment: hinton_and_le_cun_nail_in_the_coffin_knowledge_rep.pdf
Received on Sunday, 1 June 2025 00:45:00 UTC