<div dir="ltr"><div dir="ltr"><br></div><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Sun, Mar 19, 2023 at 2:01 PM spike jones via extropy-chat <<a href="mailto:extropy-chat@lists.extropy.org">extropy-chat@lists.extropy.org</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div class="msg-2360969813263073163"><div lang="EN-US" style="overflow-wrap: break-word;"><div class="m_-2360969813263073163WordSection1"><p class="MsoNormal"><u></u> <u></u></p><p class="MsoNormal"><u></u> <u></u></p><div style="border-right:none;border-bottom:none;border-left:none;border-top:1pt solid rgb(225,225,225);padding:3pt 0in 0in"><p class="MsoNormal"><b>From:</b> extropy-chat <<a href="mailto:extropy-chat-bounces@lists.extropy.org" target="_blank">extropy-chat-bounces@lists.extropy.org</a>> <b>On Behalf Of </b>Jason Resch via extropy-chat<br><b>…</b><u></u><u></u></p></div><div><div><div><div><p class="MsoNormal"><u></u> <u></u></p></div><div><p class="MsoNormal">>…We see recurring themes of information, recursion, computation, and machines and logic. I think these are likely key to any formal definition of consciousness. …Jason<u></u><u></u></p><p class="MsoNormal"><u></u> <u></u></p><p class="MsoNormal"><u></u> <u></u></p><p class="MsoNormal"><u></u> <u></u></p><p class="MsoNormal">Jason, there is a reason I stopped worrying in the past coupla weeks that ChatGPT was going to cause the singularity.  I am a big Hofstader fan, read Escher Godel Bach twice, cover to cover, invested a lot of time into that marvelous work.  He convinced me that machine consciousness (or any other sentience or self-awareness) requires a type of recursion.  Hofstadter goes on at length about recursion and self-reference, the importance of Godel’s work to understanding ourselves.</p></div></div></div></div></div></div></div></blockquote><div><br></div><div>My friend with access to GPT-4 asked it: "Does your neural network contain such reflexive loops, or is it strictly feed forward?", below is its reply:</div><div><br></div></div><blockquote style="margin:0 0 0 40px;border:none;padding:0px"><div class="gmail_quote"><div><p style="margin:0px;white-space:pre-wrap"><font color="#0000ff">As an AI language model, my underlying architecture is based on the Transformer model, which is primarily feedforward in nature but includes some elements of recurrence or reflexivity, specifically in the self-attention mechanism.</font></p></div></div><div class="gmail_quote"><div><p style="margin:0px;white-space:pre-wrap"><font color="#0000ff"><br></font></p></div></div><div class="gmail_quote"><div><p style="margin:0px;white-space:pre-wrap"><font color="#0000ff">The Transformer model consists of a multi-layer feedforward neural network with self-attention mechanisms that allow the model to learn complex relationships between input and output sequences. The self-attention mechanism weighs the importance of different input elements relative to each other, effectively capturing long-range dependencies and relationships within the input data. This mechanism introduces a form of reflexivity or recurrence, as it allows the model to consider the context of the entire input sequence when generating outputs.</font></p></div></div><div class="gmail_quote"><div><p style="margin:0px;white-space:pre-wrap"><font color="#0000ff"><br></font></p></div></div><div class="gmail_quote"><div><p style="margin:0px;white-space:pre-wrap"><font color="#0000ff">However, it is important to note that the self-attention mechanism in the Transformer model is not equivalent to traditional recurrent neural networks (RNNs), which maintain a hidden state that is updated at each time step. The Transformer model processes input sequences in parallel rather than sequentially, which makes it fundamentally different from RNNs.</font></p></div></div><div class="gmail_quote"><div><p style="margin:0px;white-space:pre-wrap"><font color="#0000ff"><br></font></p></div></div><div class="gmail_quote"><div><p style="margin:0px;white-space:pre-wrap"><font color="#0000ff">In summary, while my neural network architecture is primarily feedforward, it includes some elements of reflexivity in the form of self-attention mechanisms that allow the model to capture complex relationships within input sequences.</font></p></div></div></blockquote><div class="gmail_quote"><div><br></div><div>Is this enough to meet Hofsteader's requirements of recursion? I cannot say.</div><div><br></div><div> </div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div class="msg-2360969813263073163"><div lang="EN-US" style="overflow-wrap: break-word;"><div class="m_-2360969813263073163WordSection1"><div><div><div><div><p class="MsoNormal"><u></u><u></u></p><p class="MsoNormal"><u></u> <u></u></p><p class="MsoNormal">I tried to convince myself that two or more ChatGPTs could train each other on their own time, which is a form of recursion and self-reference, and that process could perhaps spring into a human-level AGI with a will, with self-awareness, of all the stuff we think of as us.<u></u><u></u></p><p class="MsoNormal"><u></u> <u></u></p><p class="MsoNormal">Now after studying GPT^2 discussions and GPT^3 discussions, they all seem to devolve to nothing.  The technology I think is still coming for that process: two or more AIs training each other using background compute cycles, but I now don’t think ChatGPT is that technology or is capable of it.<u></u><u></u></p><p class="MsoNormal"><u></u> <u></u></p><p class="MsoNormal">If you know of examples of GPT-GPT discussions or GPT-any chatbot that became interesting, do share please.</p></div></div></div></div></div></div></div></blockquote><div><br></div><div>There was a paper or video I saw recently that was about how a major advance had been made by using AI models to generate prompts and rate answers to help train other models. This was seen as a major advance as previously human judges had to do all this work. I cannot find the reference at this point in time.</div><div> </div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div class="msg-2360969813263073163"><div lang="EN-US" style="overflow-wrap: break-word;"><div class="m_-2360969813263073163WordSection1"><div><div><div><div><p class="MsoNormal"><u></u><u></u></p><p class="MsoNormal"><u></u> <u></u></p><p class="MsoNormal">That belief was behind my comment last week that ChatGPT is not really thinking, but rather is working language models.  <u></u><u></u></p><p class="MsoNormal"><u></u> <u></u></p><p class="MsoNormal">I currently don’t think ChatGPT is the technology capable of causing the singularity.  I am losing no sleep, not one minute of sleep over ChatGPT.</p></div></div></div></div></div></div></div></blockquote><div><br></div><div>Whether or not GPT-3 or GPT-4 triggers the singularity, I agree neither will, but we are now in the runaway phase of recursive self-improvement by AIs, using AI to design hardware, to write software, and train itself and other AIs, it won't be long at this rate.</div><div> </div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div class="msg-2360969813263073163"><div lang="EN-US" style="overflow-wrap: break-word;"><div class="m_-2360969813263073163WordSection1"><div><div><div><div><p class="MsoNormal"><u></u><u></u></p><p class="MsoNormal"><u></u> <u></u></p><p class="MsoNormal">Oops, partially retract that last comment but in a good way: I am losing some sleep over ChatGPT, by staying up late to goof with it.  It is the coolest software tool to come along in a long time.<u></u><u></u></p><p class="MsoNormal"><u></u> </p></div></div></div></div></div></div></div></blockquote><div><br></div><div>I agree.  It is fascinating.</div><div><br></div><div>Jason</div></div></div>