Since November 2022, OpenAI’s ChatGPT has garnered much attention and has become the topic of discussion for the tech community worldwide. ChatGPT has more knowledge than any human has ever known and can converse about various topics coherently.
ChatGPT is capable of holding a cogent conversation about any topic, generating songs, poems, essays, digital photos, drawings, and animations. However, concerns about the rapid development of generative AI models, specifically large language models (LLMs), have also been raised. As a result, governments are pondering over new regulations, and prominent voices are calling for a pause in the development of artificial intelligence, fearing that it may somehow run out of control, damage or even destroy human society.
While there is excitement surrounding this new AI technology, there is also deep concern about the speed of its development. Many people worry about the existential threat it poses to humanity, and governments worldwide are considering new regulations.
Origins and Capabilities of AI Software
The contemporary explosion of AI software began in the early 2010s when a software technique called “deep learning” became popular. Using vast datasets and powerful computers running neural networks on GPUs, deep learning dramatically improved computers’ abilities to recognize images, process audio, and play games. A GPU (Graphics Processing Unit) is a type of computer hardware that can greatly accelerate the processing of AI and machine learning tasks.
However, neural networks were usually embedded in software with broader functionality, and non-coders rarely interacted with AIs directly. But ChatGPT now allows users to converse with AI directly, providing a kind of intellectual vertigo caused by the sudden improvement of software to perform tasks that were previously exclusively in the domain of human intelligence.
Working of Large Language Models (LLMs)
Despite the feeling of magic, an LLM is, in reality, a giant exercise in statistics. The language of the query is first converted from words into a representative set of numbers, after which the tokens are assigned equivalent definitions by placing them into a “meaning space” where words with similar meanings are located in nearby areas.
The LLM then deploys its “attention network” to make connections between different parts of the prompt. An LLM must learn these associations from scratch during its training phase. Over billions of training runs, its attention network slowly encodes the structure of the language it sees as numbers (called “weights”) within its neural network. If it understands language at all, an LLM only does so statistically.
The response is initiated once the prompt has been processed, and the LLM generates the text based on the input it receives. The more text the model takes in, the more context it can see, and the better its answers will be. However, longer inputs require more computing power.
Limits to LLMs
LLMs have some limitations. For example, they require significant computational resources to operate effectively. They are also vulnerable to malicious use, such as generating fake news and phishing emails. Moreover, an LLM can only generate text based on what it has been trained on and cannot learn from new experiences.
Despite their impressive abilities, LLMs like ChatGPT have significant limitations. One of the most significant is their lack of common sense. They don’t have the broad understanding of the world and context that humans do, which can lead to errors and misunderstandings. For example, they may generate nonsensical responses to prompts that require background knowledge or cultural understanding. Additionally, they may reproduce biases that exist in the data they were trained on, which could perpetuate societal inequalities.
GPT-3 can process a maximum of 2,048 tokens at a time, which is around the length of a long article in The Economist. GPT-4, by contrast, can handle inputs up to 32,000 tokens long, a novella. The more text the model can take in, the more context it can see, and the better its answers will be. However, the required computation rises non-linearly with the length of the input, meaning that slightly longer inputs need much more computing power. Therefore, there is a limit to how much data the model can take in, and its growth is not infinite.
There is also the issue of explainability. While LLMs can generate responses to prompts, they do so in a black box manner, meaning that it can be difficult to understand how the model arrived at a particular response. This lack of transparency could be problematic in contexts like law or medicine, where it is important to be able to understand the reasoning behind a decision.
Concerns Surrounding Generative AI Models
The rapid development of generative AI models like ChatGPT has raised concerns about their potential impact on society. For instance, people are worried about the existential threat posed by AI and the possibility that the software could run out of control and damage or destroy human society. Additionally, AI could eliminate certain jobs, creating economic disruption.
Governments and prominent voices worldwide are calling for the development of artificial intelligence to be paused until the ethical, legal, and social implications are better understood. This technology is still in its infancy, and there are many unknowns surrounding its impact on society.
What does the future hold?
The future of LLMs like ChatGPT is uncertain, but it’s clear that they will continue to play an increasingly significant role in our lives. As the technology improves, we may see LLMs being used in a wider range of contexts, from customer service to creative writing to scientific research. However, there will also likely be continued concern about the potential risks associated with these models, including the perpetuation of biases and the potential for misuse.
As with any powerful technology, it will be important to ensure that LLMs are developed and deployed in an ethical and responsible manner. This will require collaboration between researchers, developers, policymakers, and the public to ensure that the benefits of these models are maximized while minimizing any potential harms. With careful attention and thoughtful planning, LLMs could help us to unlock new insights, create new works of art, and solve complex problems.
Generative AI models like ChatGPT have the potential to revolutionize how people interact with computers. However, the rapid development of AI technology has also raised concerns about its impact on society. Governments and individuals worldwide must weigh the benefits and risks of this technology carefully to ensure that its development aligns with the greater good.
DELTA Data Protection & Compliance, Inc. Academy & Consulting – The DELTA NEWS – Visit: delta-compliance.com
Image: DELTA/DALL-E
