There has been a lot of AI news during the last two years. But will AI deliver on all its promises? Are we in a Generative AI hype?
New technologies often follow a predictable cycle. I have personally observed the hype cycle of new technologies before.
2000, I attended the Borland/Inprise conference in Denver, Colorado. One of the speakers, Scott McNealy from Sun Microsystems, envisioned a future where Java would be ubiquitous. Fast-forward to today, and while Java (and even older languages like COBOL) has been helpful and remains in use, it is no longer at the forefront of programming languages. It has never delivered as much as promised.
This phenomenon is called the Gartner Hype Cycle, a model that describes new technologies’ adoption, maturity, and social application. The cycle consists of five phases:
- Technology Trigger
- Peak of Inflated Expectations
- Trough of Disillusionment
- Slope of Enlightenment
- Plateau of Productivity
What is Generative AI?
Generative AI is a type of artificial intelligence that can, based on a prompt, perform specific tasks, such as generating new material statistically derived from the input data provided by the user. Both input and output can consist of text and images or even video. At their core, generative AI models are based on a so-called Large Language Model. These generative models can based on input, creating realistic answers.
Generative models are trained on large amounts of data, often text and images, from the web and other sources. One source of training data is the Common Crawl, a repository of 250 billion web pages. Generative Artificial Intelligence technology is groundbreaking because it can be controlled using natural language and used in virtual assistants to analyse and process new data.
Generative AI combines several AI systems, including Deep Learning (a Neural Network with many layers), Reinforcement learning and other Machine Learning models. Natural language processing using different types of AI has been around for a long time. Still, generative AI has accelerated enormously in the last two years since OpenAI released its first version of ChatGPT.
Generative AI Hype: Riding the Wave of the Hype Cycle
As of August 2023, Gartner positioned Generative AI at the “Peak of Inflated Expectations” phase. This stage is characterised by high enthusiasm and inflated expectations about the technology’s potential (Pure AI). Despite significant investments and promising advancements, some experts predict that Generative AI may enter the “Trough of Disillusionment” within 2-5 years. This phase occurs when initial enthusiasm wanes as early implementations fail to deliver on overhyped promises.
The Reality Check: AI’s Current Capabilities
While groundbreaking, the technology is currently underdelivering in several areas. Enterprises are beginning to realise that AI can’t solve every problem. Here are a few sobering truths about AI’s capabilities, which indicate that we may be entering a Generative AI hype:
- Job Replacement: AI is unlikely to take over most jobs soon. It may automate specific tasks but will only partially replace human roles.
- Code Writing: AI can assist in writing code but can only contribute to 10-30% of the process. Human intervention remains crucial to meeting expectations or achieving high code quality.
- Marketing Copy: AI tools can generate content but need more nuanced understanding to create compelling and contextually appropriate marketing materials consistently. If you don’t edit the content well, there will be errors, or users will recognise certain words, such as “delve”. Medical researchers suddenly started using the word “delve” four times more often after the advent of ChatGPT. There are many places where AI-generated text is not allowed. That does not mean someone cannot use Generative AI for professional marketing copy; you can, but it has to be fact-checked and rewritten extensively. It can save time but hardly replace humans.
- Artificial General Intelligence (AGI): The dream of AGI, an AI with human-like reasoning and understanding, remains distant. Current AI models, including large language models (LLMs), excel at pattern recognition but need more authentic learning or reasoning abilities.
Other examples
Generative Adversarial Networks (GANs) and foundation models like those used in AI powered image generators and AI-generated content are at the forefront of current generative AI technologies. These models enable the generation of text and images for content creation across industries. While the ability to generate text or produce highly realistic images is impressive, it’s essential to remember that these innovations are still maturing. As with other technologies, we must be cautious not to let inflated expectations overshadow the practical value they offer today.
The Funding Bubble: A Cautionary Tale
Since Sam Altman at OpenAI presented ChatGPT in November 2022, an enormous amount of money has been invested in new Generative AI companies, sparking concerns about a potential bubble. Investment in AI is probably only one in the rows of bubbles in promising technologies. This is not the first time; it is typical for any new technology and part of what is expected based on the Technology hype cycle. Venture capitalists and investors are pouring billions into AI startups, hoping to capitalise on the next big technological breakthrough.
However, history has shown that such frenzied investment often leads to inflated valuations and unrealistic expectations. As with previous tech bubbles, there is a risk that many of these investments will not yield the anticipated returns, leading to significant financial losses and market corrections.
The Value of AI: Augmenting, Not Replacing Humans
Despite these limitations and acknowledging that the Generative AI hype exists, AI still holds significant value when used to augment human capabilities rather than replace them. Here are some practical applications of AI that demonstrate its potential:
- Complementary Tool: Companies are finding success using generative AI as a complementary tool. It can handle repetitive tasks, provide data-driven insights, and assist decision-making processes.
- Strategic Use: AI should be deployed thoughtfully and strategically. Businesses can enhance productivity and innovation by focusing on areas where AI can provide clear benefits.
- Workflow Integration: AI is becoming integral to workflows, complementing other technologies. It can streamline processes, reduce manual workload, and improve efficiency across various domains.
With this in mind, Generative AI will deliver enormous productivity gains and financial value. But like all hyped technologies, being more humble about the predictions is wise.
Embracing the Future of AI
The journey of AI, particularly Generative AI, illustrates the cyclical nature of technological evolution. This hype cycle does not look different from earlier hype cycles. So, while it may only fulfil some expectations, its potential to enhance and augment human efforts is undeniable. By tempering expectations and strategically integrating AI into our workflows, we can harness its power to drive progress and innovation. Ultimately Effective AI will emerge, but until then the road may be bumpy.
Navigating the Generative AI hype cycle
As we navigate the Generative AI hype cycle, it’s crucial to remain realistic about AI’s capabilities while being open to its transformative possibilities. By doing so, we can ensure that AI becomes a valuable ally in our quest for technological advancement. As with other contemporary technologies such as BlockChain, Virtual Reality, and 3D printing, which are also likely to underdeliver compared to the hype, Generative AI will be helpful but will not deliver on the promises.
Since the general principles of new technology are nothing new, studying the Gartner Hype Cycle may be worthwhile. While we should follow the development of the hype cycle for Generative AI, we may see that it will not deliver as much as promised. There are plenty of opportunities to use the new technology for many applications. There are strong reasons to believe that Generative AI is entering the “Peak of Inflated Expectations” phase. It doesn’t mean the technology is useless; it may not be as ubiquitous as promised.
Gislen Software and Generative AI
While it is realistic to assess that Artificial Intelligence is presently oversold and overhyped, based on the Generative AI hype cycle, we believe we should still use the technology as much as possible and recommend it to our clients. Just because there is hype doesn’t mean it is not valuable. Like all technologies, it can deliver good value but rarely solve all problems. Contact us to discuss how we can help you with AI or any software development.