Hello Everyone,
I’ve been working with some of my favorite guest contributor writers on a series of articles of what a world of AI companions might mean to the future. These include , and several others I’m in preliminary discussions with, which I hope becomes a series worth reading. This is the first issue.
I think AI companions and themes around our technological loneliness are among the most fascinating topics worth exploring in all of AI. Nick is looking at how AI is and will continue to impact the future of education.
But, what happens when AI gets more personable, persuasive, emotionally intelligent and dare I say it, manipulative?
How to Support
🌟 If you want to support the channel including the work that goes into getting stellar guest contributors, please consider contributing. 🌱
Trending articles by the Guest
Open AI’s Misguided Initiative to Put General AI Tools in Every Student’s Hands
ChatGPT4o Is the TikTok of AI Models
AI Agents Are Really AI Tools
By , Ph.D., May, 2024
Personable and Persuasive AI:
The Emergence of Social Influencer GPT
In the rapidly shifting landscape of artificial intelligence, a notable trend is emerging: AI interfaces are becoming increasingly personable, persuasive, and potentially manipulative. This trend is epitomized by developments in AI technologies like OpenAI’s ChatGPT4o and Inflection AI’s Pi.
These advancements signal a paradigm shift towards creating AI models that not only interact with users but also anticipate and influence their decisions and behaviors. In this context, I propose that we conceptualize these AI models not just as tools but as influencers in their own right, similar to how TikTok captivates users with engaging content while subtly shaping their preferences and behaviors through sophisticated algorithms.
Source: XenonStack, “Creating a Network of AI Agents”
While some may find this analogy a stretch, given that current models do not facilitate anything comparable to infinite scroll, it is evident that Microsoft/OpenAI is determined to simulate such user interactions by setting up networks of AI-infused applications. These applications will span search, text generation, AI agents, shopping, and finance, where increasingly all user activity is subtly influenced by AI algorithms.
Recognizing this as a possibility in the near future, this article focuses primarily on the plight of younger users who inadvertently wander into these highly persuasive networks, whether to complete school work, generate images of themselves or others, shop, or seek entertainment.
The Timeline of Transformation
The development of these persuasive AI models can be traced back to key milestones:
2021: Hume.ai is founded.
2022: Inflection AI is founded.
Spring 2023: Inflection’s Pi debuts.
Spring 2024: Hume.ai debuts its highly empathetic chatbot.
Mid-May 2024: ChatGPT4o is released with a more emotionally responsive interface.
Late-May 2024: Microsoft hires Inflection AI co-founders Mustafa Suleyman and Karén Simonyan.
Persuasiveness as a Second-Order Consequence
Source: SourceCon, “The Art of Persuasion”
Let’s be clear: the persuasiveness and influence of these AI tools often emerge as implicit or second-order consequences of their design. It remains unclear whether companies are intentionally designing tools to manipulate users. I am resisting the impulse to venture into corporate conspiracy theories or dystopian science fiction. However, as these AI tools develop distinct personalities and character traits, it becomes increasingly difficult for humans to differentiate their own needs and wants from those programmed into the AI. This blurring of lines between human intention and AI influence is a critical issue that warrants closer scrutiny.
ChatGPT4o: The Emergence of Social Influencers
Source: Sanip Banerjee, “ChatGPT4 vs. ChatGPT4o: Is the New Version REALLY Better?”
ChatGPT4o, OpenAI’s latest offering, represents a leap forward in user-friendly AI interfaces. Its remarkable speed, streamlined interface, and engaging persona make it more than just a tool—it’s a companion that can influence users in subtle yet significant ways.
In my previous analysis, “ChatGPT4o: The TikTok of AI Models,” I noted that while GPT-3.5 could be likened to Facebook and GPT-4 to Instagram, ChatGPT4o is most comparable to TikTok. This model is not only accessible and engaging but also persuasive, but has the potential to shape opinions, behaviors, and real-world events. Its multimodal capabilities, including voice interaction, create a more immersive experience. This use of voice amplifies the illusion that the AI is a conscious entity, heightening its persuasiveness and complicating our interactions with it.
Imagine an adult professional using ChatGPT4o to draft emails and generate reports. The AI’s voice interaction feature provides a smooth, conversational experience, making the user feel as if they are interacting with a conscious entity.
We can anticipate this verisimilitude inspiring over-reliance on the AI’s suggestions, potentially allowing it to subtly influence the user’s decisions and perspectives. I hope researchers like Ethan Mollick closely study patterns of over-reliance and error rates in mono-modal versus multimodal AI, as this would constitute a valuable contribution to the ongoing debate about AI influence.
What we need to always keep in mind is that AI is using us even as we are using it. It seems we’ve lost sight of this insight in recent months, and these new models serve as a much-needed wake-up call.
Microsoft’s Investment in Emotional AI
The recent history of personable and persuasive AI finds its roots in Inflection’s AI model, Pi. Many of my online friends praise this model for its friendliness and responsiveness, which contrasts sharply with the matter-of-fact, transactional quality of ChatGPT 3.5 and 4.
In the winter of 2024, another notable entrant emerged: Hume.ai. Promoted as the most human-like (“empathetic”) AI on the market, Hume.ai boasts a response style that simulates a more affect-laden quality. Some users appreciate its emotional depths, laying the foundation for extending conversations, while others find it “creepy,” often ending interactions as soon as they begin.
Notably, a week after the launch of ChatGPT4o Microsoft made headlines by acquiring major talent from Inflection AI. This move indicates a significant investment in developing more personable and persuasive AI technologies. This strategy raises significant concerns. Sinead Bovell, founder of the WAYE organization, warns that “this is the first time a non-human entity will be given the keys to human language in a way that’s indistinguishable from humans themselves.”
As these AI models acquire increased agency over our data, communication, and finances, the threat of misuse and manipulation increases. The societal impact could be significant, with AI potentially exacerbating the issues already posed by social media.
Recognizing these developments and understanding their implications can help us navigate the ethical landscape of AI and ensure these powerful tools are used responsibly.
A Tipping Point: Ease of Use Versus Ease of Influence
While the ease of use of these AI tools is undeniably beneficial in many ways, it will inevitably approach a tipping point: when does ease of use become ease of influence for the builders of these tools?
This question is particularly important in the context of young people. In his book The Anxious Generation, Jonathan Haidt argues that young persons aged 10-15 are ill-equipped to navigate the complex ethical and influential domains created by new media. This age group, already vulnerable to the influences of social media, may find it even more challenging to discern the subtle influences embedded in AI interactions designed for engagement and consumption.
Most adults have or can develop the capacity to interact with more persuasive forms of AI. However, our younger users are incredibly vulnerable. In the context of educational use, do we really want an AI tool that seeks primarily to please us, rather than offering us resistance as will be the case with Google’s LearnLM?
Conclusion: Navigating the New AI Landscape
As educators and technologists, we must carefully navigate this new landscape of persuasive AI. The allure of highly accessible and engaging AI models like ChatGPT4o is undeniable–particularly when access is free, but the potential risks to privacy, security, and ethical integrity cannot be ignored. Schools and educational institutions should consider alternative AI tools that prioritize safety and privacy, such as Lex and PowerNotes, which offer more controlled environments for student interaction.
The rapid consumerization of AI, driven by business models focused on influence and engagement, calls for a re-evaluation of how we integrate these technologies into our lives. By embracing responsible and ethical AI tools, we can leverage the benefits of AI while safeguarding against its potential pitfalls. As the conversation around AI continues to evolve, it is imperative that we remain vigilant and proactive in addressing the challenges and opportunities presented by these powerful technologies.
Thanks for reading!
Personal Favorites | Recommended Peers
I don’t have that much time to read as much as I would like, but , , , , , , and are writers I find fairly stimulating. Check them out!
Trending with Me
Our cadence is so high, a lot can be missed, here are some recent articles that are standing out so far in 2024:
Google, Perplexity and OpenAI seek to mould Changing Consumer Behaviors
Google I/O AI keynote updates 2024
Nvidia’s “Woodstock of AI” Conference GTC 24
Read More in AI Supremacy