Mainstreaming generative AI offers current capabilities, promises of future advancements, and some pitfalls.
This form of artificial intelligence technology gained momentum with the release of OpenAI’s ChatGPT and Dall-E 2 in 2022. These tools, and successors such as Google’s Bard, will make high-quality AI-generated content an everyday reality. Each of these products is built on language models, a type of machine learning model that is honed on large amounts of training data.
AI models are now available to users for a variety of content creation tasks. Text, images, video, audio, and synthetic data are all mixed together. However, the potential benefits of this technology come at some cost. Here are seven generative AI challenges business he leaders should consider.
1. Dealing with technical complexity
Generative AI models can have billions, even trillions of parameters, making them a complex undertaking for a typical business.
Arun Chandrasekaran, Vice President, Analyst, Technology Innovation, Gartner, said: Due to the need for resources, this technology can be expensive and unfriendly, so near-future deployments will see companies tap into generative AI via cloud APIs to limit coordination. He says he is likely to.
The difficulty of creating models leads to another problem. The concentration of power in a few well-financed entities, Chandrasekaran added.
2. Dealing with legacy systems
Incorporating generative AI into an outdated technology landscape can pose additional challenges for businesses. IT leaders are faced with the decision of whether to consolidate or replace older systems.
For example, financial institutions considering how to use language models to determine fraud will find new technology at odds with how current systems handle that task, says consulting firm West Monroe. Partner Pablo Alejo said.
Traditional systems “have very specific ways of doing that, and now we have generative AI that leverages different types of thinking,” Alejo explains. “Organizations must create integrations or find new ways to adopt new capabilities using new technologies so that they can achieve the same output or results more quickly and effectively.”
Generative AI technology touches across key facets of the enterprise.
3. Avoid technical debt
Generative AI can add technical debt to legacy systems if companies fail to achieve significant change through its adoption.
Companies deploying AI models for customer support may declare optimization victories because human agents will have fewer cases to handle. But according to Bill Bragg, his CIO at enterprise AI SaaS provider SymphonyAI, the workload reduction is not enough. He noted that the number of agents in frontline support roles would need to be significantly reduced to justify investments in AI.
“How did you optimize if you don’t remove anything?” said Bragg. “All you did was add more debt to the process.”
4. Reorganize part of the workforce
Generative AI is likely to reshape the way work is done in many areas, raising unemployment concerns. An article on China’s video game industry says artists’ job opportunities are disappearing as companies adopt AI-based image generators.
If you don’t have a team working on how to understand [generative AI’s] If you take advantage of the feature and take advantage of it, you run the risk of becoming obsolete.
Pablo AlejoPartner of West Monroe
But some executives suggest it’s not all doom and gloom. In his customer support example, AI could reduce the number of agents, but technology will also create other roles, Bragg said. Businesses will need AI-powered staff to monitor and improve customer experiences, he reasoned. The employee who answered the customer’s question on the slow laptop instead drives the next data and technology improvement. Bragg calls this transition the “doer-to-trainer transition.”
Similarly, Alejo said that while generative AI will eliminate some types of jobs, it “opens up entirely new types of jobs that the same people can leverage.”
5. Misuse Potential and Monitoring of AI Hallucinations
AI models reduce the cost of content creation. Not only does this help businesses, but it also helps threat actors who can more easily modify existing content to create deep fakes. Digitally altered media closely mimics the original media and can be hyper-personalized. “This includes everything from voice and video spoofing to fake his art to targeted attacks,” Chandrasekaran said.
Attackers can exploit generative AI systems, but the models themselves can mislead users. AI hallucinations provide misinformation and make up “facts.” Chandrasekaran added that depending on the domain, he could hallucinate 10% to 20% of the AI tool’s responses.
6. Monitor legal issues and algorithmic bias
New technologies can also run into intellectual property issues and expose your business to legal action. “Generative AI models have the added risk of requiring large training data without considering author approval, which can lead to copyright issues,” said Chandrasekaran. .
Algorithmic bias is another source of legal risk. Generative AI models systematically produce biased results when trained on flawed, incomplete, or unrepresentative data. AI bias can spread unchecked throughout the system, affecting decision makers who rely on results, leading to discrimination.
Chandrasekaran says flawed AI models “can propagate biases downstream in the data set, and homogenization of such models can lead to single points of failure.” .
7. Provide Coordination and Monitoring
New technologies often require organizations to launch centers of excellence (CoEs) to focus on effective recruitment and deployment. Such centers could play an important role in generative AI.
“Without a team that understands this feature and is working on how to take advantage of it, it risks becoming obsolete,” warns Alejo. “Centers of Excellence should exist in every industry and every organization.”
Such specialized groups can also create policies to govern acceptable uses of generative AI. “The CoE should take the lead in designing policies and making decisions about how different individuals across the organization can use it,” he advised Alejo. The center should seek reviews and input from key stakeholders, including legal, IT, risk and potentially other departments such as marketing, human resources and research and development, he added.
