Fact vs. Fiction: The True Scope of AI Advancements and Startup Potential

The skyrocketing success of ChatGPT has inspired a wave of AI startups popping up left and right. Some folks are raising eyebrows, worrying that this might be a red flag. But when you take a closer look, the situation is actually pretty straightforward. Generative AI isn't as mind-blowingly advanced as we might think, and there are plenty of innovation fields out there that don't need billions in venture capital to make a splash and stand their ground. I’ve taken a short effort to write a quick rundown of these areas and the derived main innovation areas. 

1. Text (maturity: 8 of 10): Generative AI models have made significant progress in natural language understanding and generation. Models like GPT-4 and its successors are capable of composing coherent and contextually relevant responses, but there is still room for improvement in understanding complex and nuanced language. 
2. Photos (5 of 10): Generative models like DALL-E & Midjourney have demonstrated impressive capabilities in generating novel images from textual descriptions. However, there is still progress to be made in terms of overall consistency, quality, and coherence of the generated images. We’re not there yet. 
3. Video (3 of 10): Video generation is more challenging due to the increased complexity and computational requirements. There have been advances in video synthesis and manipulation, but generating high-quality and coherent videos remains an area of active research and is likely years away. 
4. Music (3 of 10): AI-generated music, such as that produced by OpenAI's Jukebox, has shown promising results in creating background music and simple compositions. However, generating more complex and emotionally resonant music that rivals human compositions is still a challenge. 
5. Voice Conversations (4 of 10): Text-to-speech and voice synthesis technology has improved significantly, but there is still a gap in achieving truly natural-sounding and emotionally expressive AI-generated voices. Continued research in this area is expected.
6. Multimodality (2 of 10): Integrating multiple modes of communication, like text, images, and videos, is a challenging task for AI and is better solved by other algorithmic disciplines. While there have been some initial efforts, achieving seamless and coherent multimodal interactions remains a long-term goal for AI research.

Numerous AI startups tend to concentrate on one of these areas:

1. Specialized text applications built on top of large language models like OpenAI's GPT or other similar models, with or without additional training. While these startups have the potential to succeed, maintaining a competitive edge in terms of technical innovation could be challenging in the long run.
2. Focused applications for image, video, or music generation. These startups can achieve significant success if they can bridge the gap between existing limitations in foundational technology and the desired level of quality. This area will also require significant investments with high risks and returns. 
3. Multimodal services and applications: This area bears a strong resemblance to the current state of innovation. By combining AI, IT, and service engineering, startups can address and solve problems that were previously unresolved or inadequately addressed. This is where most venture capital is likely to be invested and will demonstrate a more or less “normal” success curve. 

Hope this made some sense. Feel free to disagree, contribute or DM.