Category: Top News

IBL News | New York

Yale University announced that it will invest over $150 million to support AI development and AI literacy among students, faculty, and staff. This investment will include computing infrastructure, access to secure generative AI tools, targeted faculty hires, seed grants, and opportunities for interdisciplinary collaboration.

More specifically, the investment will be distributed on:

1. Research Infrastructure: Yale will acquire approximately 450 GPUs over the next few years, enhancing its high-performance computing capabilities for AI research.

2. Secure AI Access: The university is launching the Clarity platform, offering a safe (walled-garden environment) that hosts an AI chatbot powered by OpenAI’s GPT-4o for the Yale community.

3. Faculty Expansion: Yale plans to recruit over 20 new faculty members specializing in AI-related research and teaching.

4. Curriculum Development: Seed grants will be offered to review and adapt curricula across various disciplines in the context of AI.

5. Collaboration Initiatives: Yale will host AI-focused research symposiums, prompt-a-thons, and seminars and offer seed grants to foster interdisciplinary collaboration.

In a message to the university community, Yale Provost Scott Strobel explained that the $150 million commitment responds to the report of the Yale Task Force on Artificial Intelligence, elaborated by the 18-member group of faculty and campus leaders.

Strobel said that the Clarity platform is appropriate for use with all data types, including high-risk data, while all security standards are observed.

Its chatbot can create content, provide coding assistance, perform data and image analysis, provide text-to-speech, and more. Over time, the platform may expand to incorporate additional AI tools, including other large language models.

“Clarity is designed to evolve as generative AI develops and the community offers feedback.”

In addition to Clarity, the university will offer faculty, students, and staff access to other AI tools, including Microsoft Copilot Enterprise and Adobe Firefly.

IBL News | New York

Microsoft launched a dozen Copilot-branded products this week, such as Pages and Agents. In addition, the software giant announced an improved Copilot in Microsoft 365 apps, especially in Excel, PowerPoint, and Outlook.

Microsoft Copilot is built into its Bing search engine, Windows 10 and 11, Edge sidebar, and iOS and Android apps. Fine-tuned versions of OpenAI’s models power it.

• Copilot Pages is a canvas designed for multiplayer collaboration. It will be generally available later in September 2024. Users can try it at Microsoft.com/copilot.

• Copilot in Excel, generally available, added new features, like Excel with Python.

• Copilot Agents makes it easier to automate and execute business processes.

These are the videos with Microsoft Copilot reported enhancements:

 

 

IBL News | New York

Researchers from the University of Chicago presented experimental evidence in a paper that ChatGPT excels at investing tasks, including predicting corporate investment policies, processing dense corporate disclosures, and detecting corporate risk.

Beyond investing, experts predict that ChatGPT will disrupt many high-skilled occupations, including journalism, IT support, human resources, and marketing.

In collaboration with Statistics Denmark, the authors surveyed over 100,000 workers from 11 occupations exposed to ChatGPT between November 2023 and January 2024.

Half of workers have used ChatGPT, with adoption rates ranging from 79% for software developers to 34% for financial advisors.

Workers differ in their intensity of ChatGPT usage, with 32% currently using it and 6% having a Plus subscription.

The survey saw substantial productivity potential in ChatGPT, estimating that it can halve working times in a third of their job tasks.

Researchers also discovered a staggering gender gap in ChatGPT adoption: Women are 20 percentage points less likely to use ChatGPT than men in the same occupation.

IBL News | New York

The new OpenAI’s o1 reasoning model —released on Thursday— scored 83% in a qualifying exam for the International Mathematics Olympiad (IMO), while GPT-4o correctly solved only 13% of problems.

Also, the coding abilities were evaluated in contests, and as indicated in an OpenAI research post, they reached the 89th percentile on competitive programming questions (Codeforces) competitions.

Trained with reinforcement learning to perform complex reasoning, this new LLM that excels in math and coding thinks before it answers—it can produce a long internal chain of thought before responding to the user.

According to OpenAI, it performs similarly to PhD students on challenging benchmark tasks in physics, chemistry, and biology.

“Our large-scale reinforcement learning algorithm teaches the model to think productively using its chain of thought in a highly data-efficient training process. We have found that o1’s performance consistently improves with more reinforcement learning (train-time compute) and with more time spent thinking (test-time compute). The constraints on scaling this approach differ substantially from those of LLM pretraining, and we are continuing to investigate them,” said the company.

o1 is rate-limited; weekly limits are currently 30 messages for o1-preview and 50 for o1-mini.

An additional downside is its high, expensive price. In the API, o1-preview is $15 per 1 million input tokens and $60 per 1 million output tokens. That’s 3x the cost versus GPT-4o for input and 4x for output. (1 million tokens is equivalent to around 750,000 words.)

OpenAI says it plans to offer o1-mini access to all free users of ChatGPT but hasn’t set a release date. We’ll hold the company to that date.

IBL News | New York

Open AI announced yesterday the release of a new model called o1 that excels in math and coding. This model is trained to answer more complex questions and solve more problems in science, coding, and math.

For example, o1 can be used by healthcare researchers to annotate cell sequencing data, physicists to generate complicated mathematical formulas needed for quantum optics, and developers in all fields to build and execute multi-step workflows.

The o1 beta can be tested in ChatGPT Plus.

“We trained these models to spend more time thinking through problems before they respond, much like a person would. They learn to refine their thinking process through training, try different strategies, and recognize their mistakes,” explained the company.

The o1 model is being released alongside the o1-mini, a smaller, cheaper version optimized for STEM reasoning.

These models aren’t available yet in the API.

We’re releasing a preview of OpenAI o1—a new series of AI models designed to spend more time thinking before they respond.

These models can reason through complex tasks and solve harder problems than previous models in science, coding, and math. https://t.co/peKzzKX1bu

— OpenAI (@OpenAI) September 12, 2024

IBL News | New York

Harvard University unveiled a study conducted in the fall of 2023 showing that student engagement in the classroom doubled when an AI Tutor was tailored to a physics course.

This tool helped students learn more material. They also self-reported significantly more engagement and motivation to learn.

These preliminary findings have inspired other large Harvard classes to test their approach, The Harvard Gazette wrote. More specifically, the Derek Bok Center for Teaching and Learning is collaborating with Harvard University Information Technology to pilot similar AI chatbots in a handful of large introductory courses this fall.

Harvard University is also developing resources to enable any instructor to integrate tutor bots into their courses.

This custom-designed AI chatbot proved surprisingly more effective than a typical “active learning” classroom setting in which students learn from a human instructor as a group.

The study was led by lecturer Gregory Kestin and senior lecturer Kelly Miller. They analyzed the learning outcomes of 194 students enrolled last fall in Kestin’s Physical Sciences 2 course, which is physics for life sciences majors.

“We went into the study extremely curious about whether our AI tutor could be as effective as in-person instructors,” Kestin, who also serves as associate director of science education, said. “And I certainly didn’t expect students to find the AI-powered lesson more engaging.”

“It was shocking and super exciting,” Miller said, considering that PS2 is already “very, very well taught.”

The researchers wrote in their paper that the experiment shows the advantage of using AI tutoring as students’ first substantial introduction to challenging material. “If AI can be used to effectively teach introductory material to students outside of class, this would allow “precious class time” to be spent developing higher-order skills, such as advanced problem-solving, project-based learning, and group work.”

However, Kestin and Miller warned about potential misuses:

“AI tutors shouldn’t ‘think’ for students but help them build critical thinking skills. AI tutors shouldn’t replace in-person instruction, but help all students better prepare for it — and possibly in a more engaging way than ever before.”

“Students with a very strong background in the material may be less engaged, and they’re sometimes bored,” Miller said. “Students who don’t have the background sometimes struggle to keep up. So the fact that this AI tutor can support that difference is probably the biggest thing.” 

The website that hosts the tutor was built on ChatGPT. Still, rather than defaulting to ChatGPT behavior, the custom tutor provided users with information guided by research-based and refined prompt engineering and “scaffolding’ to ensure the lessons were accurate and well-structured.

Mathematics instructor Eva Politou will introduce a version of this AI tutor to Math 21a (Multivariable Calculus) this fall. Every week, students will generate questions about a specific topic and search for answers with the AI tutor as a guide.

“The primary goal of the AI tutor is to promote an inquiry-based studying method,” Politou explained. “We want students to practice generating questions, critically approaching real-life scenarios, and becoming active agents of their own understanding and learning.”

 

 

 

IBL News | New York

Dartmouth College and Nvidia’s Deep Learning Institute (DLI) launched a new Generative AI Teaching Kit to provide students access to tools, frameworks, and practical exercises.

According to Joe Bungo, Deep Learning Institute (DLI) program manager at NVIDIA, “this teaching kit equips students with a deep understanding of generative AI techniques and enables educators to foster future innovation and creativity in the industry.”

IBL News | New York

OpenAI, which has 1,700 employees, is making substantial changes to its management team and corporate structure to attract large investors in a deal that would value it at $100 billion.

However, it has not yet settled on a new structure.

According to an elaborate report from The New York Times, the transition has been difficult, and early employees continue to leave.

“Rapid growth hasn’t resolved a fundamental question of what OpenAI is supposed to be: Is it a cutting-edge A.I. lab created for the benefit of humanity, or an aspiring industry giant dedicated to profits?” wrote the paper.

Only three remain of the 13 people who helped found OpenAI in late 2015 with a mission to create artificial general intelligence (A.G.I). One, Greg Brockman, the company’s president, has taken a leave of absence through the end of the year, citing the need for time off after nearly a decade of work.

According to a person familiar with OpenAI’s income, its annual revenues have topped $2 billion. ChatGPT has more than 200 million weekly users—twice the number from nine months ago—and more than 1 million paid users for the corporate versions of ChatGPT.

San Francisco-based research lab said Thursday that the figure includes the total number of people signed up to use its ChatGPT Team and Enterprise services, which are aimed at companies and universities using its ChatGPT Edu product.

The company said just under half of OpenAI’s corporate users are based in the US. Outside the US, the company’s chatbot is most popular with business users in Germany, Japan, and the United Kingdom.

However, it is unclear how much the company spends yearly, though one source puts the figure at $7 billion. Microsoft, already OpenAI’s largest investor, earmarked $13 billion for the AI company.

As some employees departed, they were asked to sign legal papers that said they would lose their OpenAI shares if they spoke out against the company. This incited new concerns among the staff, even after the company revoked the practice.