NUNZIUM

The year 2023 was a whirlwind in the artificial intelligence (AI) landscape, characterized by numerous product launches, setbacks, and strides towards regulation. AI giants Meta, Google, Baidu, OpenAI, and the French open-source contender, Mistral, all launched new AI applications. However, none of these became an immediate success, illustrating the complex nature of AI development.

AI-powered search engines from Microsoft and Google fell short of expectations, revealing fundamental flaws in language models, the backbone of AI applications. Microsoft’s Bing was embroiled in controversy for suggesting conspiracy theories, while Google’s Bard was found to generate incorrect answers, leading to a significant $100 billion loss in Google's share price.

The darker aspects of AI also came to the fore in 2023. Generative AI models exhibited severe gender, ethnic, and political biases. The environmental impact of AI was also highlighted, with research showing that generating an image using a powerful AI model consumes as much energy as fully charging a smartphone.

Despite these challenges, significant progress was made. OpenAI and Google started allowing companies and developers to create customized AI chatbots and build applications using AI. Google DeepMind utilized a large language model to solve a long-standing scientific problem, and a new open-source system, Dobb-E, was developed to teach robots simple household tasks.

The potential risks associated with AI sparked a heated debate among scientists, business leaders, and policymakers. While some expressed grave concerns, others, including Yann LeCun, Joelle Pineau, and Joy Buolamwini, dismissed these fears as distractions from the current issues caused by AI.

AI policy and regulation took center stage in the US Senate and the G7. In December, European lawmakers passed the AI Act, introducing binding rules for the development of high-risk AI. The White House introduced an executive order on AI and received voluntary commitments from leading AI companies for more transparency and standards. A proposal for watermarks in AI-generated content also gained traction.

The year saw a record number of lawsuits filed by artists and writers against AI companies for scraping their intellectual property without consent or compensation. OpenAI's superalignment team, led by Ilya Sutskever, focused on preventing a superintelligence from going rogue.

In response to an increase in copyright disputes related to AI, there was a call to designate 2024 as the Year of AI Regulation in the U.S., emphasizing the need for comprehensive accountability by governing bodies.

Looking ahead to 2024, AI regulation, especially the impact of generative AI on privacy, securities, and antitrust laws, will be a dominant topic. While there is an expectation for incremental regulatory steps as new AI applications emerge, concerns remain that premature regulation could slow progress in AI development.

Different countries have different approaches to AI regulation. The US, UK, EU, and China are all trying to strike a balance between innovation and regulation. The Biden administration released an executive order promoting safe and secure AI development, while the EU's proposed AI Act classifies AI applications into four risk categories. China requires companies to undergo security assessments and receive clearance before releasing AI products to the public.

Despite progress in generative AI, artificial general intelligence remains a distant goal. Generative AI's ability to generate text and images presents both significant benefits and challenges, such as the production of false information.

As we move into 2024, the hope is for a balanced regulation that mitigates potential harm without overly restricting beneficial technology. The need for independent regulation of AI safety and public release of AI systems has been underscored by the upheaval at OpenAI. The journey on the AI rollercoaster is just beginning, and the destination is yet to be determined.