What is the current state of GenAI in legal tech, and what are the biggest barriers to adoption? Tom M. covers the latest state of the market in our blogpost along with our approach to dealing with hallucinations and the issue of trustworthiness. https://lnkd.in/dbWHReVU
关于我们
We help firms deploy LLMs in a safer manner, reducing hallucinations and increasing trust in GenAI applications.
- 网站
-
https://www.truthsystems.ai/
Truth Systems的外部链接
- 所属行业
- 科技、信息和网络
- 规模
- 2-10 人
- 类型
- 私人持股
- 创立
- 2023
Truth Systems员工
动态
-
Excited to announce Gateway - a leading hallucination detection pipeline meant for enterprises ?? Gateway: - outperforms all existing solutions on general QA hallucination detection. - runs in real-time (avg. 0.80s per query on a NVIDIA A10). - is deployable on-premises, and locally fine-tunable. - returns citations for all predictions. See the image below for a benchmark on HaluEvalQA, a common hallucination detection dataset for question and answering. We fine-tuned our pipeline with 16,000 examples and tested Gateway (and other models) on the remaining 4,000 examples. Gateway scores the highest in accuracy (92.2%) and F1 score (0.922), and outperforms GPT-4o by 4.5% in accuracy and 4.2 F1 points. It is currently the most performant solution benchmarked. ???????? ?????????? ?????????????? ??????????????? LLMs perform poorly at language tasks like natural language inference, while small language models struggle with limited adaptability and world knowledge understanding. Gateway is an optimized, hybrid pipeline, composed of a series of both large and small language models. Gateway leverages the vast understanding of LLMs to supplement highly performant SLMs, achieving an easily fine-tunable and performant hallucination detection pipeline. ?????????? ????????? We are beginning to roll out Gateway to our existing customers in law, and will release fine-tuned versions for law-specific tasks. If you're interested in chatting more about how we made Gateway or how you can use it, feel free to reach out!
-
-
Our Co-Founder Nam Nguyen and Product Manager Tom M. will be at ILTACON 2024 from 11th-15th of August. We would love to catch up with you if you are attentding this year’s conference if you are working on GenAI applications in the legal tech space!
-
-
“In his most recent year-end report on the state of the federal judiciary, Chief Justice [John] Roberts cautioned that the ‘use of AI requires caution and humility,’” wrote Newsom. “ I wholeheartedly agree. Importantly, though, I also agree with what I take to be the report’s assumption that AI is here to stay. Now, it seems to me, is the time to figure out how to use it profitably and responsibly.“ Bloomberg Law wrote an interesting article highlighting the potential use case and risk involved with using AI in legal tech. Part of the concern is in the potential for hallucinations, but the benefits such as quick summarization and sourcing through documents proves to be powerful. What do you think is the role of GenAi in the legal industry? Truth Systems has developed Gateway, a retrieval augmented generation LLM guardrail to help legal tech firms tackle?hallucinations. Feel free to reach out to us if you are interested in developing safe LLMs! https://lnkd.in/dHNrmPK4
-
Check out our founder Nam Nguyen thoughts on the elections and what it means for GenAI https://lnkd.in/ds_P9r7V
Co-Founder @Truth Systems | Making LLM Trustworthy with Real-Time Hallucination Detection
???????????? ???????????????? ?????? ???? ?????? ???????? – ???????? ???????? ?????????? ?????? ?????????? ???? ?????? ???? ?? So, it really did happen on a casual Sunday afternoon. We heard the backroom conversations and the public outcries, but that didn’t make this news any less shocking. Biden’s unexpected move now opens the door to numerous questions and speculations about America's future, and one I’m certainly paying attention to is the future of Generative AI. Keeping track of policies and regulatory opinions on AI has been a key part of my job at Truth Systems, with the goal of helping firms build more compliant and trustworthy AI, so here is what Biden dropping out means for generative AI in the US: 1. ?? ?????? ???????????? ???? ??????????’?? ???? ?????????????????? ??????????: The future of AI will be shaped by the next administration's approach to regulations, funding, and research. In his recent platform adopted by the RNC on Monday, Trump promised to repeal Biden’s AI Executive Order with a fully hands-off approach. The Democratic nominee, mostly likely to be Kamala Harris, will surely need a clear strategy, either continuing Biden’s efforts or easing regulations. 2. ?? ?????????? ?????????? ??????????: It is no surprise that semiconductor stocks absolutely tanked last week after Trump’s comment on Taiwan. As AI relies on powerful chips like oxygen to do complex computational tasks, a disruption to Taiwanese chip manufacturing would put the entire tech industry into chaos. Eyes will certainly be on the White House regarding chip fabrication to protect or resolve this supply chain bottleneck. 3. ???????????????? ?????? ????????????????????: Changes in immigration policies and educational funding will affect the flow of international talent into the US, which is vital for maintaining the country's edge in AI research and development. During the Trump Administration, the H1-B denial rate rose to 24% in 2018 and fell to 2% in 2022 during the Biden administration. With Canada and England loosening their visa process for highly skilled workers, the new White House leadership will need to step up if it does not want to face a possible scenario of a brain drain. 4. ???????????????????????? ????, ???? ?????? ???????? ??????????????: The Biden administration has initiated discussions on AI ethics and safety, laying down a Blueprint for the AI Bill of Rights. The continuation, enhancement, or rollback of these frameworks will significantly shape how the race toward AGI will look like. Deregulation would mean a step backward in building more transparent and trustworthy AI, while overregulation would completely stifle innovations. As we navigate this political shift, staying informed and adaptable will be key. The landscape is changing, and so are the opportunities and challenges ahead. ??What do you think this holds for the future of AI? Leave a comment down below, or let me know what topics you want me to cover in my next posts.
-