For the study, the researchers focused on a type of GenAI model known as a transformer, which forms the backbone of LLMs like GPT-4.
Researchers have found out that with all the progress claimed to be made by large language models , generative artificial intelligence has still got a lot to learn and they are still incapable of being trusted fully.This is especially because a LLM that seems to be performing well in one context might break down if the task or environment slightly changes.
However, when the researchers closed some streets and added detours, the LLM’s navigation performance deteriorated significantly. senior author Ashesh Rambachan, assistant professor of economics and a principal investigator in the MIT Laboratory for Information and Decision Systems .Study focused on a specific type of GenAI model
In another example, they found out that a transformer could predict valid moves in a game of Connect 4 nearly every time without understanding any of the rules.metrics that can test a transformer’s world model. The researchers focused their evaluations on a class of problems called deterministic finite automations .
Generative AI Harvard LLM MIT Technology
United States Latest News, United States Headlines
Similar News:You can also read news stories similar to this one that we have collected from other news sources.
Meta is 'best early GenAI winner': Analysts stick by tech giant despite spending concernsAnalysts think there's still room for Meta shares to run, even after the company posted disappointing metrics in its latest quarterly report.
Read more »
California Law On GenAI Continues To AdvanceJohn Werner has created a career out of bringing ideas, networks and people together to generate powerful results. John is a Managing Director and Partner at Link Ventures. John's deep curiosity and penchant for problem-solving led him to a diverse set of roles spanning many fields and interests.
Read more »
Adapting GenAI for the next generation of learningA new study by learning analytics researchers presents key considerations for generative AI (GenAI) educational tools so they are carefully developed to support, rather than replace, human learning.
Read more »
Regulated Sectors Embrace GenAI with Caution: Trust is KeyWhile professionals in highly regulated sectors see the potential of generative AI (GenAI) to boost efficiency and innovation, concerns about safety, security, and responsible use remain. Trust is paramount as organizations seek solutions that leverage GenAI's capabilities while safeguarding sensitive data.
Read more »
Webinar: The Importance of Professional-Grade GenAI in Regulated Sectors - SPONSOR CONTENT FROM THOMSON REUTERSSponsor content from Thomson Reuters.
Read more »
3 key findings from Morgan Stanley's survey on GenAI corporate adoption3 key findings from Morgan Stanley's survey on GenAI corporate adoption
Read more »