Exploring GPT-4o, the latest generative artificial intelligence model from OpenAI
mail@perryartreed.com
ChatGPT: A “Turing” Point for Artificial Intelligence
When I began my graduate studies in 2017, it felt like artificial intelligence (AI) was a blip on the radar of public discourse, and perhaps even niche within the field of media psychology. Cut to November 2022, when the public release of ChatGPT (OpenAI, 2022) thrust AI into the public forefront. Overnight, decades of advancements seemed to culminate in a moment of widespread recognition and interest. Billions of dollars were quickly invested into further developing generative AI models (Tong, 2023) as the “next big thing” in the tech industry.
Why did AI suddenly enter, or perhaps re-enter, the zeitgeist? AI had already become an invisible force, much like electricity, powering our tech-driven lives. Most of the apps on our smartphones, for instance, rely on machine learning and other forms of AI. However, with the release of ChatGPT, AI moved from behind the scenes to center stage, capturing public attention. What changed?
My answer is that ChatGPT, with its chatbot interface, delivered a captivating user experience reminiscent of the Turing Test (Turing, 1950). In his seminal work, Alan Turing argued that the question of machine intelligence hinges not on internal mechanisms, but rather on outward displays of intelligent behavior. Turing proposed a test where a human participant interacts with a text-based computer interface and must determine if they are communicating with a machine or another human based on the text responses. If the machine’s responses are indistinguishable from those of a human, Turing (1950) argued that this would demonstrate intelligence.
Jump ahead 72 years, and ChatGPT passed the Turing test—although perhaps too well. I remember my first interactions with ChatGPT, when I was in awe of the human-like responses and ability to carry on a conversation, no different than if I were texting a friend or colleague. But by the time I used ChatGPT to generate fictional short stories for my dissertation research (Reed, 2023), I knew that something was different—ChatGPT sometimes felt smarter, faster, and debatably more creative than human-generated responses. Almost “super human.” It was an uncanny valley of human interaction. The impressive, conversational nature of ChatGPT was a “Turing” point (pardon the pun) in the modern discourse on AI.
As media psychologists, it is critical that we keep up with the advances in AI models, especially as they capture public attention. We must continue to investigate the human-computer interactions and the perceptions underlying AI, particularly with generative AI models that provide engaging user interfaces and generate novel products. As our Division’s own Sean Thoennes (2018) argues with his neo-anthropogenic psychology approach, we must take a human-based, collaborative approach to the development of AI, which has a resounding impact on individuals and societies.
Hello, GPT-4o
GPT-4 Omni (GPT-4o) was released in May 2024 and can be accessed through ChatGPT (OpenAI, 2024). Notably, the latest model makes user interactions with ChatGPT feel even more conversational. Users can input text, audio, still images, and/or videos; the AI model can generate text, audio, and still image outputs. As OpenAI (2024) notes, GPT-4o can process audio input in milliseconds, comparable to typical human response time in human-to-human interactions.
Notably, GPT-4o has been making headlines and spurring further debates over the ethics and humanistic issues of generative AI. One buzzy headline was Scarlett Johansson—who voiced an AI assistant in the 2013 film Her—alleging that OpenAI copied her voice for the new GPT-4o voice “Sky,” after she had previously declined OpenAI’s offers to use her voice (Fowler, 2024). Although OpenAI provided evidence that another voice actress was used, the intended resemblance to Johansson was difficult to dispute and they temporarily removed the “Sky” voice.
I hope these debates and discussions continue, and I hope that media psychologists play a big role in them. While developments in AI must be collaborative, cross-disciplinary efforts, we are specially trained to understand the dynamic interplays between people and machines. Given the vast influence and future potential of AI, we have our work cut out for us.
References
Fowler, G. A. (2024, May 22). The voice of Scarlett Johansson coming from your chatbot? This AI firm is making it happen. The Washington Post.https://www.washingtonpost.com/technology/2024/05/22/openai-scarlett-johansson-chatgpt-ai-voice/
OpenAI. (2022, November). ChatGPT: Optimizing language models for dialogue. OpenAI. https://openai.com/blog/chatgpt
OpenAI. (2024, May 13). Hello GPT-4o. OpenAI. https://openai.com/index/hello-gpt-4o/
Reed, P. A. (2023). Is ChatGPT creative? Cognitive-affective responses to AI-generated stories (Publication No. 30637157) [Doctoral dissertation, Fielding Graduate University]. ProQuest Dissertations and Theses Global.
Thoennes, S. (2018). Neo-anthropogenic Psychology and AI. In Proceedings of the 2018 International Conference on Artificial Intelligence.
Tong, A. (2023, February 17). Microsoft-backed OpenAI to let users customize ChatGPT. Reuters. Retrieved February 22, 2023, from https://www.reuters.com/technology/microsoft-backed-openai-let-users-customize-chatgpt-2023-02-17/
Turing, A. M. (1950). Computing machinery and intelligence. Mind, 59(236), 433-460. https://doi.org/10.1093/mind/LIX.236.433
