Large law firms are using a tool made by OpenAI to research and write legal documents. What could go wrong?
David Wakeling, head of London-based law firm Allen & Overy’s markets innovation group, first came across law-focused generative AI tool Harvey in September 2022. He approached OpenAI, the system’s developer, to run a small experiment. A handful of his firm’s lawyers would use the system to answer simple questions about the law, draft documents, and take first passes at messages to clients.
The trial started small, Wakeling says, but soon ballooned. Around 3,500 workers across the company’s 43 offices ended up using the tool, asking it around 40,000 queries in total. The law firm has now entered into a partnership to use the AI tool more widely across the company, though Wakeling declined to say how much the agreement was worth. According to Harvey, one in four at Allen & Overy’s team of lawyers now uses the AI platform every day, with 80 percent using it once a month or more. Other large law firms are starting to adopt the platform too, the company says. Read More
Daily Archives: February 24, 2023
New research suggests that privacy in the metaverse might be impossible
A new paper from the University of California Berkeley reveals that privacy may be impossible in the metaverse without innovative new safeguards to protect users.
Led by graduate researcher Vivek Nair, the recently released study was conducted at the Center for Responsible Decentralized Intelligence (RDI) and involved the largest dataset of user interactions in virtual reality (VR) that has ever been analyzed for privacy risks. Read More
AI is Useful for Capitalists but Probably Terrible for Anyone Else
AI is finally useful for business, and everyone is likely underestimating its impact. But unless AI is open-source and truly owned by the end users the future for everyone but the software providers looks grim.
The last time your author opined about the state of artificial intelligence I predicted that commercial success required two things: first, that AI researchers focus on solving a specific business problem, and second, that enough data exists for that specific business problem. The premise for this prediction was that researchers needed to develop an intuition of the business process involved so they could encode that intuition into their models. In other words, that a general-purpose solution would not crack every business problem. This might have been true temporarily, but it’s doomed to be wrong more permanently. I missed a reoccurring pattern in the history of AI: that eventually enough computational power wins. In the same way chess-playing engines that tried to encode heuristics about the game eventually lost to models that had enough computational power, these AI models for “specific business problems” have all just lost to the hundred billion parameters of GPT-3.
I am not known for being overly bullish on technology, but I struggle to think of everyday sorts of business examples where such a large language model would not do well. It is true that in the above example the model did terribly on questions requiring basic arithmetic (converting rent per square foot per month to rent per square metre per year, for example), but these limitations are missing the point. Computers are known to be adequate arithmetic-performing machines (hence the name), and surely future models would correct this and other deficiencies. Artificial intelligence is now generally useful for business, and I am probably not thinking broadly enough about where it will end up.
One decent guess, however, might be augmented intelligence – the idea that AI is best deployed as a tool to increase the power and productivity of human operators rather than replace them. Read More
ChatGPT for Robotics: Design Principles and Model Abilities
This paper presents an experimental study regarding the use of OpenAI’s ChatGPT [1] for robotics applications. We outline a strategy that combines design principles for prompt engineering and the creation of a high-level function library which allows ChatGPT to adapt to different robotics tasks, simulators, and form factors. We focus our evaluations on the effectiveness of different prompt engineering techniques and dialog strategies towards the execution of various types of robotics tasks. We explore ChatGPT’s ability to use free-form dialog, parse XML tags, and to synthesize code, in addition to the use of task-specific prompting functions and closed-loop reasoning through dialogues. Our study encompasses a range of tasks within the robotics domain, from basic logical, geometrical, and mathematical reasoning all the way to complex domains such as aerial navigation, manipulation, and embodied agents. We show that ChatGPT can be effective at solving several of such tasks, while allowing users to interact with it primarily via natural language instructions. In addition to these studies, we introduce an open-sourced research tool called PromptCraft, which contains a platform where researchers can collaboratively upload and vote on examples of good prompting schemes for robotics applications, as well as a sample robotics simulator with ChatGPT integration, making it easier for users to get started with using ChatGPT for robotics. Read More
I Made an AI Clone of Myself
I spent a day recording videos in front of a green screen and reading all types of scripts to create a digital clone of myself that can say anything I want her to using a platform called Synthesia.
In November, a company called Synthesia emailed Motherboard and offered “an exclusive date with your AI twin.”
“Hello, ever thought about creating your own digital twin? You’ve been invited to Synthesia’s New York studio to build your own virtual avatar, like me!” an AI clone of Synthesia spokesperson Laura Morelli said in a video embedded in the email. “Don’t miss out on learning more about the new sexy sector. Lock in your one-hour slot now to build your own avatar with Synthesia. Hurry now because spots are limited and filling up fast.” Read More
Rick's Cafe AI 