The AI industry is heavily investing in data centers and hardware, particularly NVIDIA's AI chips, to support these demanding models. Despite significant investments, challenges like hallucinations and cybersecurity issues persist. OpenAI plans to follow GPT-4.5 with GPT-5, which Altman describes as a system integrating various technologies, but its potential to achieve artificial general intelligence remains uncertain.
Key takeaways:
- OpenAI's new GPT-4.5 model is extremely expensive, costing $75 per million tokens, and is resource-intensive, leading to GPU shortages.
- The model is not a reasoning model and is described as having a unique kind of intelligence, focusing on engaging in warm, intuitive conversations.
- Despite being OpenAI's largest LLM, GPT-4.5 improves computational efficiency by more than 10x compared to GPT-4, but its performance is reportedly below other models on most evaluations.
- OpenAI plans to follow up with GPT-5, which aims to integrate more technology, but its ability to meet expectations and achieve artificial general intelligence is uncertain.