[Editor's note] Large language models are expected to be valuable assets for enhancing human creativity and problem solving.
Original link: https://www.stride.build/blog/the-llm-hype-curve
Do not redistribute without permission!
Author | Ako Gagarin Translator | Crescent Moon
Editor in charge | Xia Meng
Listing | CSDN (ID: CSDNnews)
In recent months, large-scale language models have become a hot word around the world, frequently hitting the headlines of major news. These complex models, like OpenAI's GPT-4 and Meta's LLaMA, have captured the imagination of researchers, developers, and the public.
However, like any transformative technology, large-scale language models experience hype, attendant volatility of expectations, and fear. At the end of 2022, Gartner released a Hype Cycle report as expectations for artificial intelligence and generative AI reached a crescendo.
With the development of new AI products exploding in less than a year following the announcement of GPT-4, where are we today on the hype curve for large language models?
What exactly are large language models?
Before discussing the hype curve, let's first introduce what a large language model actually is. Such models are a subset of generative AI optimized for generating text, specifically predicting the next word in a sentence given a cue and relevant context. These models are trained on very large datasets, with over a billion parameters, and fine-tuned by humans (or other large language models). Such models include BERT, GPT, and T5, among others.
At the end of the day, large language models are text calculators that know how to create human-understandable text from given cues.
The Hype Curve: From Excitement to Realism
Hype curves are often observed when a new technology emerges. In the early stages, there is great excitement and anticipation, driven by lofty promises and visionary predictions.
In the case of large language models, the ability to generate coherent and contextually relevant text drove the initial hype. The media reported on the amazing capabilities of these models, capturing the imaginations of countless people from all walks of life. At the same time, fears of misinterpretation of such tools have generated much controversy.
peak period of high expectations
As large language models have received increasing attention, expectations for their capabilities have swelled to unprecedented heights. People envision a future where AI-generated content will revolutionize journalism, customer service, content creation, and even personal assistants. However, at this peak stage, we must keep in mind that these models are far from perfect and have their limitations.
The trough of the bubble
After the expected peak, the actual situation of large-scale language models gradually surfaced, and thus entered a period of trough. While these models can generate impressive text or images, they also have the potential to generate inaccurate, biased, or meaningless output. Furthermore, at this stage, ethical issues surrounding AI and the potential misuse of such technologies are magnified.
As a result, enthusiasm has faded and public sentiment has tilted toward doubt and fear.
I think we're at that stage today, and we've accelerated past the peak of inflated expectations!
While many individuals and companies have leveraged this technology to create enormous value, there are only a few, and many are still in the trough of the bubble.
A Bright Period of Steady Climb
As the initial hype fades, understanding of large language models begins to feel more real. Researchers and developers are actively working on addressing the limitations and challenges associated with these models. Improvements have been made in fine-tuning techniques, data quality, and reducing bias.
The focus has shifted from high expectations to improved technologies for practical application. In the light of the steady climb, the true potential and value of large language models begins to materialize.
Large language models can't solve all problems, but can come very close. According to the Pareto principle (aka 80/20 rule, only about 20% of the factors affect 80% of the results), these tools only have a 20% probability of helping you create 80% of the value, depending on the use case. These models unleash creativity in ways never before possible between humans and machines. Not only does it speed up the ideation process, but it also removes many of the barriers to problem solving.
plateau of real production
Eventually, large language models will find their own footing and make meaningful contributions across multiple industries. Improved deployment strategies, a better understanding of one's own strengths and limitations, coupled with proper ethical considerations, can make these models valuable tools.
Large-scale language models can not only help us complete tasks such as content creation, language translation, and chatbots, but can even assist researchers in their research and development. The plateau of substantial production marks the maturation of large language models that will seamlessly fit into our lives and become tools to support them. When all this will materialize remains to be seen, but it may be sooner than we think!
Summarize
There is no doubt that large language models have caused a stir in the field of artificial intelligence. The hype curve around these models is a natural progression that any transformative technology goes through. While initial high expectations may trigger a low point, it must be acknowledged that these models have enormous potential.
As the technology continues to mature, difficult problems are overcome, and applications are improved, large language models are expected to become valuable assets for enhancing human creativity and problem solving.
Understanding and managing the hype curve can help us harness these powerful tools responsibly and use them to improve society.
