OpenAI recently unveiled its latest AI large language model (LLM), GPT-4o (Omni), and the initial reactions have been mixed. While some viewers of the short demo presentation were left wanting more, OpenAI has since released additional video demos and information about the new model. GPT-4o is faster than its predecessor, GPT-4, making it more affordable for third-party developers. One of the most notable advancements is its improved ability to detect and mimic human expressions through audio, making it more emotional.
One of the key features of GPT-4o is that it is free to use through ChatGPT for all users, even non-subscribers. However, paying subscribers will have earlier access to the update. Currently, GPT-4o only supports text and vision capabilities, with audio and video capabilities set to be added in the coming weeks. To achieve its increased speed and cost reduction, GPT-4o has been trained to treat text, audio, and visual data equally and transform it all into tokens.
The reactions to GPT-4o from various experts and influencers in the AI community further highlight the mixed response. Nvidia senior research manager Jim Fan mentioned that OpenAI’s pivot towards emotional AI with a strong personality was a departure from their previous approach. On the other hand, Pennsylvania University professor Ethan Mollick believed that GPT-4o wasn’t a significant leap forward.
Startup advisor Allie K. Miller expressed excitement about the new desktop ChatGPT app, calling it a game-changer for productivity. However, journalist James Vincent cautioned that while the marketing for GPT-4o as a voice assistant was impressive, it didn’t necessarily indicate significant advancements in capability.
Chirag Dekate, a VP at Gartner, expressed some disappointment with the GPT-4o unveiling event, comparing it to Google’s Gemini demos from three months prior. He noted a growing “capability gap” between OpenAI and other tech giants like Google and Microsoft, who have more resources and distribution channels.
Despite the mixed reactions, there were also positive responses to GPT-4o. Developer Siqi Chen praised its ability to render 3D objects from text prompts, calling the feature underrated. Greg Isenberg, the CEO of Late Checkout, remarked on the incredible pace of change that OpenAI is achieving.
Overall, GPT-4o has generated passionate responses, indicating that it has already made an impact. VentureBeat has gained access to the new model and will be testing it in the coming days, providing a more detailed analysis. As GPT-4o continues to evolve and reach a wider audience, its true potential will become clearer.
