Overview of AI A researcher has been accused of collecting materials from lecturers without their consent to train a chatbot that can solve problems on the Massachusetts Institute of Technology computer science degree assignments and exams.
controversial paper [PDF] He described training and testing GPT-4 using a dataset of 4,500 questions taken from various courses set by elite universities.
“We assess the ability of large-scale language models to meet MIT mathematics and major graduation requirements. [Electrical Engineering and Computer Science]. Our results show that GPT-3.5 successfully solves one-third of the entire MIT curriculum, while GPT-4 is rapidly engineered to achieve a perfect resolution rate on the test set, excluding image-based problems. shows to achieve. ‘ says the paper’s abstract claims.
The paper was criticized for its sloppy research methods. For example, GPT-4 already had access to the solution before it was tested and scored the answer incorrectly. Professor Armando Solarezama, who is listed as a co-author of the paper, later revealed. [PDF] Lead author Ido Drori, a research associate visiting MIT at the time, did not obtain consent to collect the dataset in the first place.
Confirmed by some teachers register That they did not share course materials with Dorori. Meanwhile, Solar-Lezema, who only helped write the paper’s foreword, said Drori argued that GPT-4 would be enough for him to graduate with a computer science degree at MIT. He said he didn’t know he was using the dataset for
The paper has since been withdrawn from arXiv, and Drori is reportedly no longer working at MIT.
Generative AI Raises Antitrust Concerns, Federal Trade Commission Says
The FTC is keeping a close eye on the generative AI hype to ensure that dominant players don’t squash the competition as the technology increasingly permeates society.
Officials warned that as the market matures, companies may seek ways to control the resources needed to develop models, including data, hardware and human resources.
They pay close attention to cloud providers who try to overwhelm their rivals by offering their own AI services at low prices or locking customers into their platforms. Big tech, on the other hand, could gain power by acquiring startups that leverage new technologies and expertise.
The FTC said that while the open-source movement could counteract some of these concerns by making it easier for developers to build their own systems, especially if the software was released to the public without safety guardrails in place. noted that the technology could be abused.
“Generative AI is still evolving rapidly, but it already has the potential to transform many markets. Researchers are free to move on to the work they do best,” the statement said. “If we advance the state of technology, entrepreneurs can continue to innovate.”
“As the competitive issue around generative AI continues to evolve, the Competition Bureau will work closely with the Technology Bureau and leverage a full range of tools to identify and address unfair competition methods.”
OpenAI office opens in London
OpenAI opens its first international office in London.
The company announced the move this week, saying it hopes to attract more researchers and engineers to work on AI advancements alongside its U.S. employees.
“We see this expansion as an opportunity to attract world-class talent and drive innovation in AGI development and policy,” OpenAI CEO Sam Altman said in a statement. . “We are excited to see what the future holds and how our London office contributes towards building and deploying secure AI.”
The UK technology industry is growing and continues to be a leader in AI, especially in research. London is also home to one of his top AI labs, DeepMind, so it makes sense for OpenAI to establish a presence in London as well. OpenAI is headquartered in San Francisco, where he reportedly employs 375 people full-time, according to reports. information. ®
