Is OpenAI’s latest model, GPT-4.5, worth the hype and the reported $70 billion investment? Early user tests paint a complex picture of a chatbot with improved emotional intelligence but surprisingly underwhelming performance on traditional benchmarks.
In a move that has left many in the AI community scratching their heads, OpenAI unveiled its newest large language model, GPT-4.5, in a pre-dawn livestream that lasted less than 15 minutes. While CEO Sam Altman, fresh from becoming a father, was absent from the presentation (reportedly tending to his newborn), he took to X (formerly Twitter) to share his initial thoughts.
Altman’s assessment was intriguing, to say the least. He described GPT-4.5 as the first model that genuinely felt like conversing with a thoughtful individual. However, he tempered expectations by acknowledging that it’s a large and expensive model and, crucially, not the kind of model that scores high on benchmarks.
This sentiment was echoed in OpenAI’s official blog post, which highlighted GPT-4.5’s enhanced emotional intelligence, broader knowledge base, and reduced tendency to hallucinate (fabricate information). Yet, the blog also conceded that the model underperforms compared to even smaller models like o3-mini on datasets like GPQA (science), AIME’24 (mathematics), and SWE-Bench Verified (coding).
The Verdict: More Feels Than Facts?
The emerging consensus, based on early user experiences, suggests that GPT-4.5 prioritizes nuanced communication and empathetic responses over raw processing power and logical reasoning. This shift raises fundamental questions about the direction of AI development. Is OpenAI betting on emotional intelligence as the key to unlocking the next level of human-computer interaction, even at the expense of traditional performance metrics?
GPU Shortage and Future Plans
Altman also addressed the elephant in the room: the scarcity of GPUs. He explained that the initial plan was to launch both Plus and Pro versions of GPT-4.5 simultaneously, but the rapid pace of development has outstripped the available hardware. OpenAI is working to acquire tens of thousands of additional GPUs in the coming weeks to address the shortage and prioritize the Plus version rollout, with plans for hundreds of thousands more in the future.
The Big Question: Is it Worth the Investment?
The news raises a critical question: Is a model that prioritizes EQ over IQ worth the massive investment OpenAI has reportedly poured into its development? While enhanced emotional intelligence could have significant implications for applications like customer service and mental health support, the underwhelming performance on core benchmarks raises concerns about its utility in more demanding fields.
Looking Ahead
The launch of GPT-4.5 marks a pivotal moment in the evolution of AI. It signals a potential shift away from a purely performance-driven approach towards a more holistic model that values emotional intelligence and nuanced communication. Whether this gamble pays off remains to be seen, but one thing is clear: OpenAI is betting big on a future where AI understands not just what we say, but how we feel.
References:
- AI好好用. (2024, February 28). 实测GPT4.5，智商一般、贵得离谱，OpenAI砸5000亿就造出这玩意？[Tested GPT-4.5, average IQ, outrageously expensive, OpenAI smashed 500 billion to make this thing?]. Machine Heart. [Original article link would be inserted here]
Views: 0