Generative Ai Models the Risks and Potential Rewards in Business · Table (p.10)
From generative-ai-models-the-risks-and-potential-rewards-in-business.pdf · page 10
| 0 | 1 |
|---|---|
| You’ve now seen how generative AI models may help | to protect its intellectual property while still being able to enjoy |
| consumers, streamline organizational processes and free | the benefits of generative AI applications. |
| up time for employees to take on higher-value organizational | |
| tasks. That said, the use of generative AI has many limits and | |
| Employee misuse and inaccuracies | |
| potential pitfalls. | |
| Even legitimate use of generative AI comes with risk. The | |
| As we mention in part 4, GPT-3.5, ChatGPT’s initial underlying | |
| models generate responses based on input received, | |
| LLM, was trained on material up to September 2021 and wasn’t | |
| meaning there’s a risk they may provide false or malicious | |
| connected to the internet. And though OpenAI has made it | |
| content. As your employees use it, they need to be cautious | |
| possible for ChatGPT to browse the internet in certain cases, | |
| and review AI-generated content with a critical eye and | |
| it’s still crucial to ensure human review and expertise is built | |
| emphasis on quality assurance. | |
| into the generative AI use process. Generative AI models can | |
| If generative AI content contains inaccuracies that are not | |
| be the core of an AI application but require additional analytics, | |
| caught, this could impact your business’ outcomes or create | |
| technology and human process around them to solve problems. | |
| liability issues. For instance, Meta’s generative AI bot Galactica | |
| In this section, we discuss the risks of using generative AI | |
| was created to condense scientific information to help | |
| models and applications and how to manage them, including | |
| academics and researchers quickly find papers and studies. | |
| around client and company confidentiality, employee misuse | |
| Instead, it produced vast amounts of misinformation that | |
| and phishing. | |
| incorrectly cited reputable scientists.15 Another Meta bot, | |
| BlenderBot3, was caught making false and biased claims16 | |
| Large multimodal models like ChatGPT generate human-like | |
| shortly after its release in August 2022. As well, Google’s | |
| responses. However, they lack human-like reasoning skills. | |
| chatbot Bard caused parent company Alphabet to lose US$100 | |
| For them to be considered trusted, users are responsible | |
| billion in market value after it shared incorrect information | |
| for applying their AI capabilities to suitable use cases, and | |
| during its first demo.17 ChatGPT hallucinating facts is also well | |
| your organization should educate employees on using such | |
| documented,18, 19 with developer OpenAI acknowledging its | |
| programs. Equally important, developers should use reliable | |
| ongoing shortcomings.20 | |
| data sets to train the AI models and apply relevant bias and | |
| content filters. | |
| Other risks around generative AI include the possibility that | |
| the technology could generate sensitive information, such as | |
| Risk management | personal data, that could be used for identity theft or to invade |
| one’s privacy. Even a disgruntled employee or angry customer | |
| Generative AI’s growing popularity is another reason we | |
| could create fake material to harm your company’s reputation | |
| recommend developing and deploying AI in a responsible way | |
| or that of one of your employees or executives. | |
| if your organization wants to protect itself against misuse. | |
| The following are some of the risk management challenges of | |
| generative AI models. | Generative AI evolves |
| As the world’s understanding of AI continues to evolve, we | |
| Internal risks and considerations | are already seeing a rising number of global regulations. It’s |
| important to stay abreast of these, even if you don’t plan on | |
| Breaking confidentiality and intellectual property | |
| using generative AI intentionally. | |
| Many generative AI models are built to absorb user-inputted | |
| We expect that generative AI will continue to be integrated into | |
| data to improve the underlying models over time, in essence | |
| many common applications, systems and processes, ranging | |
| helping them learn and build knowledge. That data, in turn, | |
| from internet browsers to AI-connected technology that your | |
| could be used to answer a prompt from someone else, | |
| organization may license. Thus, it’s key to be vigilant and make | |
| possibly exposing private or proprietary information to the | |
| sure you don’t use AI professionally in a way that contravenes | |
| public. The more your business uses this technology, the more | |
| applicable laws (including privacy laws), client agreements or | |
| likely it is others could access your sensitive or confidential | |
| professional standards. | |
| information. Thus, your organization needs to figure out how |
Source Metadata