Have you ever played the game Snake? You know the one where you control a line that grows longer as it consumes dots on the screen. The game becomes increasingly challenging as the snake grows, and you must navigate it carefully to avoid collisions with itself. Eventually, the snake becomes so large that it seems almost forced to start eating its own tail.
This seemingly simple game offers a striking metaphor for what's happening with generative AI at the moment. As AI models consume more and more data, they grow in complexity and capability. But what happens when these models start to feed on their own outputs? What if the AI, like the snake in the game, starts to consume itself?
The world is going MAD!
You may or may not be aware of the phenomenon called Model Autophagy Disorder (MAD), where AI models are trained on synthetic content generated by other AI models, leading to a form of data inbreeding. Researchers are finding out that AI models start to act weird with distorted outputs in a self-consuming loop. Fidelity is always lost when making a copy of a copy, and simply put, that is what happens when these models consume their own information. This is fast becoming more of a concern because of the vast amounts of data that are required to train these learning models. If no one is there to filter out the synthetic data from the authentic, the quality of the output of the AI is significantly reduced. This may lead to biases and misinformation rendering the technology unusable or at least to be taken much less seriously.
What will AI be in Five Years?
I’m sure that we can all agree that AI has great potential for a new kind of web experience and real-world products. But, will platforms like Open AI become the new Microsofts or Apples of the tech industry? Will models like ChatGPT become their operating systems where updates will be slow and reactionary? If so, what does that mean for the extremely fast past adoption of this new technology. Could the fast past adoption of AI be stifled if it takes too much time to update these models in affect, leaving AI to wither on the vine or…
New Markets and Opportunities
Generative AI is opening doors to new markets and opportunities. From improving customer and employee experiences to bolstering knowledge management and accelerating software delivery, AI is making its presence felt. It's enabling humans to be more productive by using AI to produce first drafts and foundational content. Research opportunities are also expanding, with AI analyzing large volumes of data to create new forms, summaries, and insights.
As we stand on the cusp of this transformation, we must ask: Was the cart put before the horse? It seems that this is a rare instance where a very technologically advanced system/tool was released to the public without intense testing and experimentation. Though this new tool is being tested and applied to many real-world issues like disease prediction and prevention, mental health, transportation and education to name a few areas. But it seems that its main adoption is still as a toy for the masses creating neat pictures and helping with homework. So, what happens when, like computers, AI becomes a tool required for everyday life in both private and public worlds and the web is already flooded with synthesized and recycled AI generations?
This may be how a new market manifests itself within nascent technology? Perhaps a new future job market might be centered on creating original content for the sole purpose of feeding these AI beasts. Entire buildings may be filled with people reviewing internet content to weed out synthesized data to keep the AI models in good health. This would involve new infrastructure for this new endeavor. New software tools for filtering, new forensic tools for investigating and identifying forgeries and counterfeits online. All of which will undoubtedly use AI to develop. And in a new way the snake continues to eat itself. Perhaps this will be the human’s ultimate role in this technology as it continues to automate application and learning… to keep the tale away from its mouth.
Conclusion
The future of AI is a fascinating and uncharted territory, filled with potential and ringing with uncertainty. As we navigate the complex landscape of generative AI, we must be mindful of both the opportunities and challenges it presents. The Snake game metaphor is more apt than ever, as we must carefully guide the growth of AI to avoid the pitfalls of self-consumption and ensure that it serves the greater good.
Will we see a new era of innovation and prosperity, or will we be stuck in an ongoing grapple with unforeseen consequences? Only time will tell, but one thing is clear: the AI revolution is here, and it's reshaping our world in ways we are only beginning to understand.
