2024人工智能报告502页

LAIArtificial IntelligenceIndex Report2024Message Fromthe Co-directorsA decade ago,the best Al systems in the world were unable to classify objects in images at a human level.Alstruggled with language comprehension and could not solve math problems.Today,Al systems routinely exceedhuman performance on standard benchmarks.Progress accelerated in 2023.New state-of-the-art systems like GPT-4,Gemini,and Claude 3 are impressivelymultimodal:They can generate fluent text in dozens of languages,process audio,and even explain memes.As Alhas improved,it has increasingly forced its way into our lives.Companies are racing to build Al-based products,and Al is increasingly being used by the general public.But current Al technology still has significant problems.Itcannot reliably deal with facts,perform complex reasoning,or explain its conclusions.Al faces two interrelated futures.First,technology continues to improve and is increasingly used,having majorconsequences for productivity and employment.It can be put to both good and bad uses.In the second future,the adoption of Al is constrained by the limitations of the technology.Regardless of which future unfolds,govemments are increasingly concerned.They are stepping in to encourage the upside,such as funding universityR&D and incentivizing private investment.Governments are also aiming to manage the potential downsides,suchas impacts on employment,privacy concerns,misinformation,and intellectual property rights.As Al rapidly evolves,the Al Index aims to help the Al community,policymakers,business leaders,journalists,andthe general public navigate this complex landscape.It provides ongoing,objective snapshots tracking severalkey areas:technical progress in Al capabilities,the community and investments driving Al development anddeployment,public opinion on current and potential future impacts,and policy measures taken to stimulate Alinnovation while managing its risks and challenges.By comprehensively monitoring the Al ecosystem,the Indexserves as an important resource for understanding this transformative technological force.On the technical front,this year's Al Index reports that the number of new large language models releasedworldwide in 2023 doubled over the previous year.Two-thirds were open-source,but the highest-performingmodels came from industry players with closed systems.Gemini Ultra became the first LLM to reach human-level performance on the Massive Multitask Language Understanding(MMLU)benchmark;performance on thebenchmark has improved by 15 percentage points since last year.Additionally,GPT-4 achieved an impressive 0.96mean win rate score on the comprehensive Holistic Evaluation of Language Models(HELM)benchmark,whichincludes MMLU among other evaluations.心