User:Darwin2049/ChatGPT4-Sandbox

(2023.04.23@16:30; sharper, more focused, succinct; shorter, cut to the chase; (2023.04.24@13:30; keep focus on deep learning; what it is, can do; maintain short, preferably bullet point descriptions for discussion purposes; relegate observations, discussions to notes, references section;

Overview. In the following we try to analyze and contextualize the current known facts surrounding the OpenAI ChatGPT4 (CG4) system.

• CG4: we offer a summary of how OpenAI describes it; put simply, what is CG4?
• Impressions: our focus then moves to examine what some voices of concern are saying;
• Impact: we then shift focus to what is being said about it; we note that in some cases people express astonishment and wonder; in other case they express fear and concern;
• Scenarios: here we focus on how CG4 might be used be used in expected and unexpected ways;

What is CG4? CG4 is a narrow artificial intelligence system, is rumored to have been trained on one trillion parameters; if this is true then it is a substantial leap forward beyond the previous 175 billion parameters of ChatGPT3; A peek at ResearchAIMultiple offers a comprehensive set of descriptions on how CG4 improves over CG3; to list just a few - it:

• has been developed using the generative predictive transformer (GPT) model;
• has been trained on a very large data set including textual material that can be found on the internet;
• is capable of sustaining conversational interaction using text based input provided by a user;
• can provide contextually relevant and consistent responses;
• can link topics in a chronologically consistent manner and refer back to them in current prompt requests;
• is a Large Language Models that uses prediction as the basis of its actions;
• uses deep learning neural networks and very large training data sets;
• uses a SAAS model; like Google Search, Youtube or Morningstar Financial;

Impressions (observations made by informed individuals)

• Some things that people have said about it:
  • possess no consciousness, sentience, intentionality, motivation, feelings or self reflectivity;
  • is a narrow artificial intelligence;
  • is available to a worldwide 24/7 audience;
  • can debug and write, correct and provide explanatory documentation to code;
  • explain its responses;
  • write music and poems;
  • summarize convoluted documents or stories;
  • score in the 90% level on the SAT, Bar and Medical Exams;
  • provide answers to homework;
  • self critiques and improves own responses;
  • provide explanations to difficult abstract questions;
  • calibrate its response style to resemble known news presenters or narrators;
  • provides convincingly accurate responses to Turing Test questions;

• Positive.
  • has demonstrated performance that suggests that it can pass the Turing Test;
  • CG4 derivative such as a CG5 could exhibit artificial general intelligence (AGI) capability;
  • emergent capabilities;
  • spatial, mathematical reasoning;
  • is capable of generating responses that are 25000 words in length compared to the 5000 of ChatGPT-3.5;
  • demonstrates better adherence to alignment imperatives;
  • is capable of engaging in a Socratic dialog

• Concerns.
  • knowledge gaps: inability to provide meaningful or intelligent responses on certain topics;
  • deception: might be capable to evade human control, replicate and devise independent agenda to pursue;
  • intentionality: possibility of agenda actions being hazardous or inimical to human welfare;
  • economic disruption: places jobs at risk because it can now perform some tasks previously defined within a job description;
  • emergence: unforeseen, uncontrollable capabilities;
  • “hallucinations”: solution, answers not grounded in real world;

Impact. There is an ongoing worldwide debate about the impact that CG4-type capabilities can have. Thus far there are voices that are cautioning against its further development. Geopolitical realities however may neutralize these concerns. Should a major nation-state demonstrate heretofore unexpected capabilities with this new technology then all other participants may find themselves in a catch-up mode.

In a recent statement, the president of Russia, Vladimir Putin noted that whoever is able to achieve the high ground in this technology race will possess a sustainable lead in affairs of state. Therefore some middle ground might need to be found to preclude a preemption on the part of one or another authoritarian regime and the risks associated with an unrestrained artificial intelligence system. That this is the case can be seen by the downside risks that have already occurred as a result of the accidental (or deliberate) release of the STUXNET intrusion tool.

An core feature of these deep learning systems is that they utilize large data sets for training purposes. Environments where large data sets are available will have an inherent advantage. At least one such case comes to mind and that is the PRC. It uses a system called WeChat for a remarkable variety of purposes. Individuals interact with each other as well as engage in commercial activities on a daily basis. A data set of this magnitude must offer a considerable advantage to any efforts that the CCP might have in mind.

Risks.. CG4 is a technology whose full import is going to take time to understand and grapple with. Like any new technology it can be used for constructive as well as destructive purposes. Based upon the reporting so far we believe that CG4 will have impact across a wide variety of communities. Individuals who have till now been gainfully employed may find that their job definitions are being either redefined at best, or pruned, at worst they may find themselves sidelined.

• Theoretical.

Some knowledgeable observers have expressed the concern that a derivative of CG4 could achieve consciousness. How this might happen presupposes that some form of emergent properties crystallize from the complexity and architecture of the system. These concerns are not without merit. A real world example of emergence can be seen by examining how termite mounds can be constructed from the activities of tends of thousands of mindless termites; these structures reveal an astonishing degree of sophistication despite there being no central architect or team of engineers; However the path to consciousness is littered with pitfalls. Just to list a few: a conscious entity would possess sentience, intentionality, a self model as an agent in the world; effectors that can manipulate objects and elements in the real world; to be sure, there have been advances whereby a robotic or android device is driven wirelessly by an advanced artificial intelligence system; yet there remains the set of other features that a consciousness would be expected to have; these would also include motivation, goal orientation intrinsic to itself and independent from sourcing from an external (human) entity;

As of this writing some of these are possible, others remain to be demonstrated; Recent efforts by [chrome-extension://efaidnbmnnnibpcajpcglclefindmkaj/https://arxiv.org/pdf/2304.03442.pdf Stanford Univ and Google] work in generative agents is suggestive of where more progress may develop. Based upon the approach that these teams used the work in this area bears close monitoring. This is because the team incorporated feedback loops in their system that can be used by the various agents to reflect upon and refine or otherwise improve their performance and behavior.

Malicious actors have become a common theme reported on in the 24/7 news cycle. We should expect that various nefarious actors will seize upon CG4's capabilities and use it to prompt further disruption. Because the range of capabilities that CG4 offers is so broad it is difficult to guess as to how individuals or groups of hackers might use this tool for creating trojan horse, backdoor or DOS as well as DDOS attacks. Generative Agents are powerfully suggestive but appear to still be in their earliest stages of development.

• Systemic.

A channel creator known as CP Gray posted a video some years ago titled Humans Need Not Apply. In this fifteen minute video he made the case that advancing artificial intelligence could result in as much as a 45% unemployment rate. He observed that during the Great Depression the unemployment rate was 25%. Moreover he concluded that large segments of recent college graduates would not only find themselves unemployed but unemployable. A review of the job categories that he mentioned in his video are already within the purview of CG4 to perform.

Malicious actors have become a common theme reported on in the 24/7 news cycle. We should expect that various nefarious actors will seize upon CG4's capabilities and use it to prompt further disruption. Because the range of capabilities that CG4 offers is so broad it is difficult to guess as to how individuals or groups of hackers might use this tool for creating trojan horse, backdoor or DOS as well as DDOS attacks.

Speculations. The range of possible areas that CG4 is very broad; in fact, society wide; therefore we will try to maintain focus by presenting a few possible scenarios of how its capabilities might be used;

• Influence, Persuasion. One can envision using CG4 to generate a narrative story. Business leaders, particularly at the executive level such as a CEO or members of a board of directors have as one of their required skills the ability to weave a narrative vision of the company's mission. This narrative is more successful when it is structured in such a way that the various stakeholders can see themselves and their roles articulated in these narratives. The skill of persuasion through storytelling is therefore vital. What CG4 has shown itself capable of is the creation of remarkable story narratives that can be endlessly honed and embellished upon.

• Novel capabilities;

Notes

References

Deep Learning Models. There are many variations on the deep learning network architecture models current today.

GPT3 (OpenAI) This system is understood to have 175 billion parameters;

Midjourney

Stable Diffusion

Generative Pre-Trained Transformers. A transformer is a deep learning model that adopts the mechanism of self-attention, differentially weighting the significance of each part of the input (which includes the recursive output) data. It is used primarily in the fields of natural language processing (NLP) and computer vision (CV).

The top ten types of deep learning models include:

• multilayer perceptrons (MLPs)
• radial basis function networks (RBFNs)
• convolutional neural networks (CNNs)
• recurrent neural networks (RNNs)
• long short-term memory networks (LSTMs)
• restricted boltzmann machines (RBMs)
• self organizing maps (SOMs)
• generative adversarial networks (GANs)
• autoencoders deep learning algorithm
• deep belief networks