Foundational AI: Opportunities & Risks Unveiled in New Report
The development of artificial intelligence (AI) presents significant opportunities for economic and social flourishing, according to a recent report. The release of foundational AI models such as the large language model (LLM) ChatGPT4 has captured the world’s attention, hinting at a transformative future in various fields including work, communication, scientific research, and diplomacy.
Goldman Sachs predicts that LLMs could potentially raise global GDP by 7 percent and increase productivity growth by 1.5 percent over the next decade. McKinsey research indicates that generative AI like ChatGPT4 could contribute up to $4.4 trillion annually across 60 different use cases, spanning customer operations, marketing, sales, software engineering, and R&D. This technology is also influencing international trade, with LLMs supporting this trend. The potential benefits of AI are immense, but it is crucial to address the risks that come with it.
One of the key risks highlighted in the report is that LLMs are trained on data that reflects existing social norms, biases, and discrimination. This poses a challenge as these models can inadvertently perpetuate these biases, leading to discriminatory outcomes. LLMs also raise concerns about information hazards, as they can provide true but harmful information, such as guidance on building explosives or committing fraud. Protecting personal privacy is another significant concern, as LLMs could inadvertently reveal sensitive information about individuals. Moreover, the misuse of LLMs could increase the incidence and effectiveness of crimes, exacerbating existing risks in areas such as misinformation and exclusion.
To mitigate these risks, governments around the world are either implementing or planning to establish regulations for AI. However, regulating AI while fostering innovation is a delicate balance, especially for a rapidly evolving technology that is still in its early stages. Effective AI governance is crucial to maximize the benefits and minimize harm. The report emphasizes the need for deeper and broader international cooperation to support domestic AI governance efforts.
Free trade agreements (FTAs) and digital economy agreements (DEAs) already include commitments to increase access to AI and enhance its governance. These commitments often focus on facilitating cross-border data flows, avoiding data localization requirements, and not mandating access to source code as a condition for market entry. However, further development and expansion of these commitments are necessary to address the opportunities and risks posed by foundational AI models like LLMs. The outcomes from international economic forums, including the G7 and the U.S.-EU Trade and Technology Council, are crucial in fostering international cooperation on AI governance. Initiatives to establish international AI standards in global standards development organizations (SDOs) like the International Organization for Standardization/International Electrotechnical Commission (ISO/IEC) are also pivotal in developing this cooperation.
By leveraging these international platforms and agreements, it is possible to develop new commitments that promote global opportunities for foundational AI models while mitigating associated risks. The report stresses the importance of utilizing these trade policies to expand the reach of foundational AI and support efforts to address risks, ultimately benefiting societies and economies worldwide.
In conclusion, the development of foundational AI models presents both immense opportunities and potential risks. While AI has the potential to revolutionize various sectors and drive economic growth, it is essential to address the biases, information hazards, privacy concerns, and criminal risks associated with these models. Striking the right balance between regulation and innovation is crucial. International cooperation through forums, agreements, and standard-setting organizations can play a vital role in achieving responsible and beneficial AI governance.