Artificial Intelligence is a topic that has been widely debated, and since OpenAI was, and I believe still is, one of the pioneering companies in this technology, it's normal that many news stories, gossip, updates, videos, and articles are precisely about the environment surrounding this leading AI company on the planet, responsible for ChatGPT, the largest AI chat.
Perhaps not everyone knows, but Chat BOT Artificial Intelligences are not as "intelligent" as the media makes them seem with news of an unimaginable future where AI will dominate the entire production chain. Is it possible? It could be, with technological advancement, this is a real possibility, but it's worth mentioning that AI chats, like ChatGPT, operate on an LLM model, known as Large Language Models. Now, what about the problems of AI, are you familiar with them? Check out some of them now!
I'll be much appreaciated if you Votes, Reblogs or give any Tips. Thanks and have good reading! 🍀
Problems with AI
These LLM models attempt to predict the next word in a sentence, based on training with a large volume of data. Therefore, the way the machine "responds" is not similar to human thought, and there are some problems with this model. The first is the need to store context; it's no coincidence that the price of RAM has risen so much with the increase in AI investment. Artificial intelligences that store context need a place to store previously discussed information, and for quick access, they use RAM.
In addition to the issue of data centers needing state-of-the-art equipment, something that not everyone considers is the lifespan of this hardware. While there is great optimism regarding the code, with the expectation that progress will increase with each new version released, on the other hand, few people comment on the enormous cost of component degradation, while the service is offered "cheaply" compared to the investment. Eventually, these components will need to be replaced, or technology will advance and something better will be created—better than running these models on existing video cards and RAM, perhaps even a new type of card.
Moving on to technology, but still talking about Data Centers, some details about them can be quite problematic. These structures are generally very technological and, although they generate a large value in tokens for AI use, there aren't many employees working there since everything is highly automated. Therefore, this isn't the type of project that will bring many jobs to a region. Furthermore, it could even end up causing losses if too many tax breaks are given, because there's a high energy consumption to keep all the equipment running, and they also say it needs a large amount of water to cool the equipment.
In short, if the project is installed in a city without infrastructure, it will increase costs for all residents, who will end up paying for the benefits given to the AI facility to be installed there, but it won't directly benefit the region, since it's a service provided globally, doesn't employ many people, and won't boost local commerce.
Considering all these details of AI's problems, and along with a CEO who is said to be difficult to deal with because he says something different to each person, we may be heading towards a future where AI is everything the industry promises, or towards a bubble, with these promises proving unattainable in the long term. The investment in this area was gigantic, perhaps even greater than that of the dot-com bubble, and at some point investors will question the financial return on their capital.
