生成型人工智能如何帮助律师改善法律服务交付

Technology has become essential to the practice of law, but the boundaries oftechnology’s impact in improving legal services delivery continue to expand.Moreand more lawyers are using artificial intelligence (AI)-based tools every day. AI is awell-established part of lawyers’ work in legal research, contract analysis, e-discovery,and predictive analysis, with tools that leverage machine learning techniques. Legal As much as those tools have revolutionized certain types of legal work, another newgeneration of AI promises an even more significant transformation of legal work and betteroutcomes for clients. The new generation is sometimes called generative AI, and its poweris based on the scale enabled by large language models or LLMs. These new tools, when Generative AI: the latest phase in a rapidly evolving It has beenhard to ignore the interestand excitement over new generative AI forms thathave recently reached the market. One of those generative AI products, ChatGPT, quicklyturned into a worldwide phenomenon. It reached a million users in less than a month, ChatGPT sprints to one million users Time it took for selected online services to reach one million users Why the interest? Why have ChatGPT and other forms of generative AI received so much interest?First, ChatGPT is freely available for anyone to test out. By simply entering prompts orcommands, it could seemingly address almost any question or execute any task that wasasked of it. Users discovered that it could effortlessly help them perform a wide variety of ChatGPT’s simple format has democratized the public’s access to machine learningtools. It takes instructions and allows users to respond with further directions in plainconversational language. This capability made it easy for people to see its value, which Generative AI has entered the collective legal consciousness quickly. Ina recent surveyconducted by the Thomson Reuters® Institute, ChatGPT and generative AI awareness issignificantly higher among legal professionals, with 91% of respondents saying they have And with that awareness comes the realization of potential use cases for using this How does it work? Why do new forms of generative AI seem to be so effective at reproducing Machine learning systems are not “thinking” systems. They are language predictionsystems. Many machine learning tools are designed to predict missing words in a sequence.When humans review those predictions for accuracy, the predictions are given a score for We have known for some time now that, as the data sets that machine learning systemsare trained on have become larger, their predictions and ability to provide accurateanswers have improved only modestly. The big breakthrough with ChatGPT was that Generative AI in legal: risks and opportunities The legal profession is in a key position relative to generative AI. First, legal work isprimarily centered around words, documents, and data. The work consists of finding, Second, legal is a field where accuracy and precision are essential. There is little room forerror when legal processes and decisions can impact parties’ important rights or financial Legal audiences are clearly impressed by the potential power of generative AI, but alsoconcerned by the dangers, as seen in the recentThomson Reuters Institute report. While ChatGPT can deliver answers that sound coherent and accurate, this type of AI isreally just a sentence-completion engine. ChatGPT has little intelligence but is very goodat drafting language that sounds like plausible responses to a user’s prompt. But it’s not Those examples are called “hallucinations” and include thenow-famous case of aNew York lawyerwho submitted a brief supporting a motion to a court that includedarguments and citations based on the results of a ChatGPT session. The brief included Errors like that present an unacceptable level of risk for legal professionals. So how willlawyers effectively use generative AI while maintaining all their client obligations? A standard methodology to improve generative AI results in specialized fields is calledRetrieval Augmented Generation (RAG). In systems that use RAG, the user’s prompts orqueries do not pass directly through to the LLM; the question runs first as a search againsta trusted body of content — for example,verified legal content from a legal publisheror trusted documents from the user’s organization. Documents relevant to the questionare retrieved first, and the question and the verified content are passed on to the LLM The power of generative AI derives from the size and scale of the original data sets thatthe AI models are trained on — but in specialized fields like law, it’s critical that training of What’s driving the adoption of AI in the A “perfect storm” of factors is pushing lawyers and legal organizations toward greateradoption of AI-based tools in their work. The main drivers of adoption include: Tasks centered on documents and data are likely candida