With the Large-Scale Adoption of Generative AI, Where is Traditional AI Heading?
-
Nearly a year after the launch of ChatGPT, companies are racing to adopt generative AI to gain a competitive edge or prevent competitors from leveraging the same technology. However, this raises the question: is there still a place for traditional forms of AI, particularly predictive models based on machine learning algorithms?
Image source: AI-generated image, licensed by Midjourney
As we step into 2023, McKinsey's latest AI report indicates that generative AI has experienced its 'breakout year,' with one-third of surveyed organizations reporting regular use of generative AI. The survey also reveals that due to advancements in generative AI, 40% of organizations plan to increase their overall investment in AI. However, interestingly, this has not led to a widespread boost in other forms of AI, particularly traditional models based on machine learning algorithms.
Data from the Fortune-Deloitte CEO survey reveals that 55% of chief executives are evaluating or experimenting with generative AI, while only 39% report doing the same for predictive AI. These figures caught the attention of Forrester analyst Kjell Carlsson, who noted substantial differences between generative AI and traditional AI.
The generative nature of AI represents the true distinction. Many companies are leveraging generative AI to develop internal assistants and chatbots based on their proprietary data, texts, and reports. Surprisingly, pharmaceutical companies are also accelerating drug discovery using generative AI.
However, compared to the rise of generative AI, the process of deploying AI applications to production has not changed much. All those traditional capabilities required for scale, integration of the latest technologies, achieving observability and transparency, and leveraging hybrid cloud for ease and cost efficiency have become even more important in the generative AI domain.
While data science platform vendors like Domino are busy adapting their business models for generative AI, there are other vendors who are more deeply capitalizing on the generative AI wave. OpenAI and its business partner Microsoft are leveraging their first-mover advantage to secure a significant share in the emerging generative AI market.
The market success of generative AI vendors reflects another important distinction between generative AI and traditional AI: generative AI is currently mainly something to purchase rather than build. This was also the topic of a recent article by John Thomas, a data and analytics consultant on LinkedIn. He pointed out that traditional AI models are mostly custom-developed, while generative AI applications are primarily built using foundation models developed by vendors.
There are other significant differences between generative AI and traditional AI projects, including that starting with generative AI requires smaller upfront development costs and can be launched within days. In contrast, traditional AI requires higher upfront costs and needs longer setup times.
Significant differences in technology, skills, costs, and data types result in varied use cases. Traditional AI is mainly used for analytical tasks, involving predicting values or classifying observations based on past data. In contrast, generative AI can generate content and perform tasks, with new capabilities including generating and manipulating code, text, images, videos, audio, and data.
As organizations rapidly transition from the experimental/evaluation phase of generative AI to limited and full-scale production, they will accumulate valuable knowledge on how to use this technology. However, as past experiences with big data, machine learning, and traditional AI have shown, the road to productivity may have unexpected twists and turns, not even considering the known issues of generative AI in terms of hallucinations, privacy, and legal liabilities.
Although the level of hype in mainstream media suggests we've achieved the ultimate goal of Artificial General Intelligence (AGI), those deeply involved in big data, advanced analytics, and artificial intelligence recognize that we still have a long way to go before achieving AGI. Additionally, given that most organizations have less than a year of experience with generative AI, the collective learning curve surrounding generative AI is inevitably steep.
During this period, generative AI will continue to capture nearly all the attention, while traditional AI will pay the price. Once the initial hype around generative AI subsides and executives realize it doesn't offer a quick and easy path to transformation success—while also introducing new challenges regarding accuracy, transparency, and legal liabilities—businesses will find more solid ground as they undertake the difficult but necessary work of integrating generative AI into existing IT stacks and business models.