Washington, September 2 — A recent listicle by Cris Tolomia, published on Quartz, sheds light on the often-overlooked facts about artificial intelligence, a technology that has been around for nearly seven decades. The article highlights the gap between the public’s perception of AI and the reality of how it works, its limitations, and its hidden costs.
The practical upshot of this is that large language models, such as ChatGPT, Gemini, and Claude, generate text by predicting the statistically most likely next word based on patterns in training data, rather than checking facts against a database. This explains why chatbots can deliver wrong answers with the same confident tone as correct ones, and why they can give different answers to the same question twice due to built-in randomness.
How AI Systems Work
Here’s what it means for users: when you interact with a chatbot, you’re not getting a straightforward answer based on verifiable facts, but rather a prediction based on patterns in the training data. The article notes that retrieval features, citations, and grounding tools are essentially patches applied to the base model because it cannot reliably distinguish true from false statements. In plain terms, these tools help to improve the accuracy of the model, but they are not a foolproof solution.
A short sentence sums it up: AI systems rely on human workers to label the data used to train them. This labor-intensive process is a crucial part of the AI development process, and it raises important questions about the role of human workers in the development of AI systems.
The article also touches on the legal implications of using copyrighted material in training data, which is an issue that courts are still resolving.
History and Limitations of AI
The field of artificial intelligence has experienced multiple boom-and-bust cycles over the years, and the current state of AI is no exception. Despite the hype and fear surrounding AI, the underlying technology is based on a relatively simple architecture: a neural network trained on text. This architecture has been the foundation of nearly every major chatbot released since 2017.
The article emphasizes that understanding these facts is essential to using, evaluating, or discussing AI with precision, rather than getting caught up in excitement or alarm. So what’s next for AI?
As the technology continues to evolve, it’s essential to separate fact from fiction and to understand the practical implications of AI systems. By doing so, we can harness the potential of AI to improve our lives and make informed decisions about its development and use.
Looking Ahead
As we look to the future, it’s clear that AI will continue to play a significant role in shaping our world. By understanding the facts about AI, we can work towards creating a future where AI is used to benefit society as a whole. One thing to watch is how the development of AI systems will be influenced by the ongoing debate about copyright and training data.
Another area to keep an eye on is the development of new architectures and technologies that could potentially improve the accuracy and reliability of AI systems. With the right approach, we can unlock the full potential of AI and create a brighter future for all.
