.svg){kind=small}
It has been over 2 years since ChatGPT launched to the public in November 2022. Ever since, there have been constant improvements in the GenAI large language model (LLM) space, with multiple startups and tech majors investing heavily in this advanced technology. Prominent examples include Bard (Now Gemini) from Google, Claude by Anthropic (Amazon-backed), and Microsoft’s Bing Chat (powered by ChatGPT).
We have seen these models evolve from pure text responses to now generating multiple minutes of Video from just text output. The GenAI space exploded from writing detailed novels to drawing portraits to making movies.
However, the last six months in this space have cast a trail of clouds over the otherwise bright landscape. The initial boom seems to have slowed down with every new model and update now making a lesser splash in the market than the last. ChatGPT’s initial adoption rates have been declining. Gemini’s market share dropped from 16% in January to 13% in November 2024, and Claude still has minimal adoption compared to its other two competitors.
So, what are some reasons why the initial boom is different from the enthusiasm and adoption it received in its initial years? We can argue that this is the trajectory for every disruptive technology where it undergoes quick adoption with more to a steady state, but there is more to it than that.
On-demand webinar
Automating Customer Service with Quiq: Empowering CX through Customer-Centric AI
Watch Now
Data availability
It’s common knowledge now that an AI model is as good as the data available to train it. Since its inception, ChatGPT, for example, has already used over 45 Terabytes of data, and industry experts believe that a large amount of usable data has already been consumed. Additionally, with the advent of GenAI, it’s becoming increasingly difficult to segregate real usable data from the one generated artificially.
Computational Constraints
It can take weeks, if not months, to train a subsequent model. Some estimates suggest that the initial ChatGPT 3 model expenses from electricity alone were $10-15 Million, with Gemini and ChatGPT 4 exceeding these costs by a multitude.
Environmental impact is also a concern, as data processing, storage, and hardware needs contribute significantly to global warming.
Ethical and Cultural Concerns
GenAI has become embroiled in political debates, with individual AI models often accused of bias. Moreover, the originality of AI-generated content is frequently questioned due to its reliance on vast training data. This raises a philosophical dilemma: given the extensive knowledge and experiences that shape human thought, can any idea truly be considered original?
Variable X
Apart from the data, the thing that separates Good AI from Great AI is the model—explicitly accounting for all the variables that should be a part of the answer. Billions of variables are already accounted for in the current GenAI models. However, genuinely mimicking human thinking would mean coming as close to the actual process of thinking as possible. A challenge with that would be the inability for anyone to truly understand the different external factors that affect the human mind.
Conclusion
Although GenAI has made substantial progress, its high costs and data dependency continue to pose barriers to widespread adoption. Small and medium-sized businesses may opt for existing predictive technologies as a more cost-effective solution. However, as industry leaders work to mitigate these challenges, the potential of GenAI remains promising.
Generative AI's growth presents both opportunities and challenges. But what does it mean for customer service? Discover how AI-powered solutions transform customer interactions and enable businesses to deliver faster, more personalized support.
Learn more in this on-demand webinar and explore how you can leverage AI to automate customer service and enhance customer satisfaction.
