Generative AI, a technology that has been generating a lot of hype in recent years, may not surpass humans anytime soon, according to experts. During a fireside chat at VentureBeat Transform 2023, Jeff Wong, global CIO at Ernst and Young, and Usama Fayyad, executive director of the Institute for Experiential AI at Northeastern University, discussed the current reality of generative AI.
Wong highlighted the difference between the hype surrounding new technologies and the actual capabilities they offer. While there are numerous opportunities for generative AI, Fayyad pointed out the hype surrounding the technology’s capabilities. He referred to large language models and generative AI as stochastic parrots that lack true comprehension and simply repeat what they have been programmed to say.
Although generative AI holds promise, it still heavily relies on human oversight and intervention. Fayyad warned about the potential for hallucinations and the trouble that randomness can cause in these models. Despite this, generative AI has real-world applications that can drive productivity and economic growth.
In comparison to previous applications of AI, generative AI is more practical and easier to use. Fayyad believes that the skills acquired through generative models align well with the repetitive tasks commonly performed in the knowledge economy. It has the potential to provide automation, cost savings, and acceleration.
To ensure accountability in the use of generative AI, Fayyad emphasizes the importance of governance. Establishing who is liable for problems and the implications of that liability is crucial. Organizations will be motivated to maintain accuracy and fairness if the responsibility lies with a person or legal entity rather than solely with the AI model.
Fayyad sees generative AI as a complement to humans rather than a replacement. Even if a generative AI tool can produce a legal brief, for example, the lawyer still needs to review and take responsibility for it. The same goes for code, where developers need to debug and be accountable for potential errors.
In conclusion, generative AI is viewed as a technology that augments human capabilities rather than replacing them. While the hype surrounding its potential should be tempered, experts recognize the value it brings to the knowledge economy. By understanding its limitations and using it as a decision-making tool, generative AI can contribute to automation, cost savings, and acceleration in various industries.
