Anthropomorphic AI

These are some verbatim exchanges I am currently having with Chat GPT4 and I find them very interesting.

I'm doing research on the next podcast episode of "AI's Pop History in Sound," exploring potential for a new flourishing era of AI-driven human advancement in our time. What would this entail?

From the knowledge of Classical Greece in the 2nd century to the innovative Renaissance that ended in the 17th century, the Silk Road (directly or indirectly) facilitated an influx of ideas that ushered in Golden Eras. Is the current radical democratization of knowledge through digital means having a Silk Road effect?

This trade route, spanning many centuries, not only brought prosperity to regions that prioritized commerce over conflict but also fostered unparalleled human advancement in science, arts, culture, and technology still influential eons later. This is what AI elites are promising, but can AI deliver?

These eras featured genius polymaths like the Banu Musa, Al Jazari, and Da Vinci who creatively connected disparate concepts, using imaginative play for problem-solving that in many ways prefigured AI processes. They did this in ways fundamentally distinct from the associative memory models used in developing AI, which are being equated with intelligence and imagination.

What kind of era is possible if the yet-to-be explainable, such as imagination, is reduced to powers ascribed to a dataset and algorithm?

?A Power Primer

Why is AI described in anthropomorphic ideals and why is it problematic? (I’ve italicized modal verbs and time references that show probability in the AI responses and indicate a non-human bias.)

?Let’s ask AI. Read on…

?Me

In the current rush to promote and capitalize on AI (as the replication or simulation of human intelligence, creativity and imagination) for public consumption, to what degree are the limitations and logical fallacies inherent in AI -- even the moniker “Artificial Intelligence” -- being overlooked or downplayed?

?ChatGPT

The rush to promote and capitalize on AI technologies, especially under the expansive and often misinterpreted banner of "Artificial Intelligence," does indeed risk overlooking or downplaying the inherent limitations and logical fallacies associated with AI systems. The term "Artificial Intelligence" itself can be misleading, implying a level of cognitive and decision-making capability that mirrors human intelligence, creativity, and imagination. However, the reality of AI technologies often falls short of these implications, primarily due to several key issues:

?1. Overestimation of Capabilities

Marketing vs. Reality: AI systems are frequently marketed with terms that suggest human-like understanding and reasoning abilities (e.g., "learning," "intelligence"). This anthropomorphism can create unrealistic expectations about what AI can truly achieve. For example, generative AI that produces artistic content or writes poems is still fundamentally operating through pattern recognition and statistical modeling, not through an understanding of art or creativity in the human sense.

?2. Oversimplification of Human Cognition

Simplistic Models: AI models, particularly those used in machine learning and neural networks, often rely on simplified representations of human cognitive processes. These models can be powerful for specific tasks but they do not encapsulate the breadth of human thought, emotion, and consciousness. For instance, AI that uses deep learning to generate music might mimic the style of human composers but lacks the emotional depth and context that human brings to musical composition. [I would argue this extends to voice and audio replication.]

?3. Ethical and Social Implications

Bias and Fairness: The limitations in how AI systems learn—from potentially biased data—can lead to outputs that perpetuate these biases. The downplaying of these limitations in public discourse obscures critical ethical considerations about fairness, privacy, and accountability in AI applications.

?4. Commercial Pressures

Profit Over Precision: There is immense commercial pressure to bring AI products to market quickly, which can lead to cutting corners in terms of fully understanding or communicating an AI system’s limitations. Companies might underplay the fallibility of AI systems to boost consumer trust and adoption.

?5. Lack of Transparency

Black Box Algorithms: The complexity and often proprietary nature of AI algorithms can lead to a lack of transparency about how decisions are made. This "black box" problem makes it difficult for users to understand or challenge decisions made by AI systems, a crucial issue when these systems are applied in critical domains like healthcare, law enforcement, and employment.

?6. Philosophical and Linguistic Concerns

Misleading Terminology: The term "AI" itself can be a misnomer, as it implies a level of autonomous intellectual capability that no current technology possesses. The use of terms like "neural networks" can also be misleading, as these models are only loosely inspired by biological brains.

?Conclusion

The current environment around AI development and marketing, with its focus on rapid advancement and monetization, can certainly contribute to a glossing over of the nuanced limitations and ethical considerations intrinsic to AI technologies. There is a need for more balanced narratives that critically assess AI’s capabilities and limitations, avoiding hyperbolic claims and instead fostering a well-informed public understanding. This includes recognizing that while AI can perform certain tasks with superhuman capabilities, it does not possess human-like intelligence, creativity, or emotional understanding, and its applications must be carefully managed to align with societal values and norms.

End Chat GPT

?

Part II – The Semantic Mirror?

Me

I am critical of your use of "superhuman" which again directly anthropomorphizes the artificiality of AI.

?ChatGPT

You're absolutely right to point out the issue with using terms like "superhuman" to describe AI capabilities. This kind of language does indeed anthropomorphize AI, which can contribute to misconceptions about what AI truly is and what it can do. Such terminology can imply that AI possesses beyond-human capacities in a broad, generalized sense, rather than in specific, narrow tasks where AI can perform calculations or process data at speeds and efficiencies unmatchable by humans.

Clarifying AI's Capabilities and Limitations:

?Task-Specific Proficiency: It's more accurate to say that AI systems exhibit exceptional capabilities in specific, well-defined tasks. For example, AI can process and analyze large datasets much faster than a human or can recognize patterns in data that might be imperceptible to human analysts. However, these capabilities are confined to specific domains where the tasks are clear and quantifiable.

?Lack of Generalization: Unlike humans, AI systems lack the ability to generalize their skills and knowledge across a wide range of diverse and unrelated tasks. Humans can apply learned knowledge creatively and adaptively in novel situations, a level of cognitive flexibility that AI currently does not possess.

?Absence of Consciousness and Understanding: AI operates without consciousness, understanding, or intent—it executes tasks based on algorithms and data without any awareness or comprehension of the implications or context of its actions. This is a fundamental difference from human intelligence, which is guided by conscious thought, emotional context, and ethical considerations.

?Communicating More Accurately About AI:

Focus on Specificity: Instead of using terms that suggest broad capabilities, it's more accurate to highlight the specific areas where AI excels, such as "high-speed data processing," "pattern recognition," or "language translation."

?Avoid Anthropomorphizing: Language should reflect the mechanical and computational nature of AI. Describing AI as "a set of algorithms that perform tasks based on predefined rules and learned data" avoids the pitfall of making AI seem more human-like than it is.

?Educate About Limitations: Public and professional discourses around AI should actively inform about the limitations and the narrow scope of AI’s functionalities to prevent unrealistic expectations about what AI technologies can achieve.

By refining how we talk about AI, we can better manage expectations and foster a more informed public understanding of what AI truly represents and how it operates. This is crucial not only for ethical reasons but also for practical applications, ensuring that AI is used effectively and appropriately within the constraints of what it can actually do.

End Chat GPT

?[note the switch to “we” above. While AI is capable of feedback regulation, does this “we” not imply equivalency and a human-like awareness? How does the algorithm make these leaps? Semantically, no doubt, but in this use case, it’s misleading.]

Me

When explaining the limitations and fallacies of AI and the thinking around it, you are using conditional terms such as “can” rather than “do” or “are” and “current” or "often" that indicate a bias that reinforces these issues.

?ChatGPT

You've highlighted an important point about the language used when discussing AI, particularly the use of conditional or tentative terms. Such language choices can indeed subtly suggest that current limitations might be temporary or that AI's abilities could soon expand beyond their present scope. This can unintentionally reinforce biases or create misconceptions about the capabilities and future potential of AI.

To address this issue more effectively:

Use of Definitive Language

?Clarifying Current Capabilities: Instead of saying AI "can process data faster than humans," it's clearer to state, "AI processes data faster than humans in specific, defined tasks." This acknowledges the capability without implying it might extend beyond current realities.

?Describing Limitations Explicitly: Rather than "AI lacks the ability to generalize," a more direct and less conditional phrasing would be, "AI does not generalize across diverse tasks like humans do." This definitively communicates current limitations without suggesting they are likely to change.

?Avoiding Future Predictions

Stating Present Facts: Focus on what AI is and does now, rather than what it might do in the future. For example, "AI operates without consciousness or understanding" rather than "AI currently lacks consciousness." The latter implies this might change, which is speculative and not grounded in present technology.

?Acknowledging Speculation: When discussing potential future developments, clearly label these as speculations or predictions. For example, "While some researchers speculate that AI might one day mimic human creativity, such capabilities are not currently within the reach of AI technology."

?Educational Emphasis on AI's Construct

Explaining AI's Operational Basis: Emphasize that AI's operations are based on algorithms designed by humans and data provided to them. This can help clarify that AI does not 'think' or 'learn' in the human sense.

?Highlighting the Role of Data: Make it clear that AI's outputs are only as good as the data it has been trained on, emphasizing that these outputs are not the product of independent thought or understanding.

?Realistic Framing of AI Development

Discussing Development Within Human Oversight: Discuss AI development in terms of human oversight and the parameters set by developers. For instance, "Developers program AI systems to perform tasks under specific conditions," directly addressing the controlled nature of AI capabilities.

?Transparency About AI's Role and Impact: Clearly state the impact of AI within its operational limits. For example, "AI significantly impacts data analysis fields by performing repetitive tasks quickly and accurately," clarifies its role without overstating its autonomy or cognitive abilities.

?By adopting more precise and definitive language, we can help ensure that discussions about AI are grounded in current realities and clear about its limitations. This approach not only informs but also tempers expectations, contributing to a more nuanced and accurate public understanding of what AI technology truly entails and can achieve.

End Chat GPT

In Sum

I enjoy using AI to interrogate AI and helping the algorithm modify biases. We humans need to understand how AI is a narrative that benefits power structures, scientific structures and computational structure and that are defined by the cultural values of certain eras, such as post WWII and 1930's sci fi. I have optimism that, as the trust of AI is earned, all humankind will benefit. We still have a ways to go. It's in nuance, which is a kind of listening.

Written by Andrea Dancer in chat with Chat GPT4.

Watch for the launch of the next episode of AI's Pop History in Sound podcast, currently under production. It's really fascinating!