“GPU compute and data pretraining require the deep pockets of billionaires, VCs and large corporations,” said Soumith Chintala, project leader at AI Meta, New York. he spoke to 500 people in one of the largest venues in the area. As computing and engineering become more expensive, and issues in our society rise about data law, security and social impact, companies are putting a stop to open research. Although large companies are quick to become leaders in this research, they also care about data law, security and social impact, he added.
Excitement, money, disruptive trends in AI and machine learning (ML) technology, rumors of new startups and job talks could be heard in the halls of the Messe Wien Exhibition Congress Center, a glass-fronted conference space which was filled with thousands of students, researchers, scientists, professors and industrialists. Like talks about visiting Vienna, going to jazz concerts, swimming in the Danube last week in July.
“We started as a global community of about 500 researchers 20 years ago,” said Katherine Heller, research scientist in Responsible AI at Google Research and chair of the ICML 2024 program, who attended the conference. this AI repeatedly for twenty years. . For the past few months, he has been reviewing proposals, creating programs, inviting speakers, and deciding on workshops and seminars for the week-long event. “This year we reviewed more than 10,000 papers from 12,000 authors, and selected about 2,600 to display,” he said, describing the growth of the community involved in ML as “revealing”.
Before 2022, AI technology was largely limited to research projects within university laboratories and companies. When OpenAI released ChatGPT to the public towards the end of 2022 – more of an experiment than anything else – its rapid adoption by users turned the academic field into a gold mine. Overnight, ChatGPT became a family affair and companies that had been working on artificial intelligence and a variety of big talkies smelled an opportunity – a general-purpose technology that was it can not only be used for some scientific research, but to code, search and generate artificial words, images, and videos for everyday use.
According to Bloomberg Intelligence, the size of the consumer AI market is expected to grow from $40 billion in 2022 to $1.3 trillion in 2023. The market capacity since then has included talent, money and other research in the field.
“AI research can have a huge impact on healthcare, education and climate issues and is an important research that we should not miss in India,” said computer scientist Sandeep Juneja who heads the Center for Learning and Make Decisions at Ashoka University. to produce a paper, to understand new avenues of research in this fast-moving field of science.
The relentless pursuit of profit and talent
The company’s heightened interest is pushing and emphasizing a particular type of machine learning, while pushing other types of ML research to the backburner. By default the most important terms in the research papers were large-scale linguistic models (LLM), reinforcement learning, deep learning and graph networks – all research focused concepts in building AI products and multimodal AI for profit.
“The research community is at the forefront of what’s happening technically in the AI universe,” said Sheila Gulati, chief executive and founder of Seattle-based Tola Capital, an AI trading firm, who was at the conference to find ways. the new ones. AI for investment.
One of the 10 best paper awards was given to GENIE (Generative Interactive Environments) a program developed by scientists at Google DeepMind that demonstrated technology that created an artificial environment based on text. Another award went to VideoPoet, an LLM video production from scratch created by scientists at LumaAI, Google DeepMind and Google Research.
“It takes money to do this research, to scale it up and to reach a lot of people,” said Heller, who was a researcher before moving to Google to work in health care. There is a lot of money to be gained by finding such kind of tools in the industry, but you have to connect your goals with what the profit-making company is most interested in at that time, he explained. “The industry is capitalism and researchers recognize that,” he said.
However, there is a difference in this breathtaking speed. “Ensuring that AI delivers sustainably is driving research and pressing researchers,” said Dr Sunita Sarawagi, professor and head of the foundation, Center for Machine Intelligence and Data Science, Indian Institute of Technology Bombay, of has been in the field since 2003. While it’s great that machine learning is making so much impact, there are many dangers in the uninformed use of AI, and we need to be careful,” he said, he added that due to such a high demand for AI researchers in the industry, he finds it difficult to attract students to his institute to do academic research.
For AI/ML experts are in high demand not only for tech companies and startups but also for the financial industry and academics like Sarawagi need to compete with everyone. This presented a recruiting channel for the conference. Every day, several jobs were posted for Phd students, engineers, AI scientists and more, while recruiters actively interacted with students in the hallways. “I saw some impressive people with good research and software skills,” said M Saquib Sarfraz, technical lead at Mercedes Benz Tech Innovation, based in Germany, whose sole job is to send an engineer You have already received more than 50 responses.
California-based Matthew Leavitt, co-founder and chief scientist at Datalogy AI, sponsored the conference in hopes of getting his newly-invented brand name out into the community. “The cost of supporting the conference and travel will be recovered if we hire one candidate, for that is less than what we pay the recruiter for an AI researcher,” he said, adding that there is competition difficult (their site is close to the big one. Google) but they have been able to gather interest due to the fact that they are working on the preservation of data, a “frontier research problem”, which is an important research objective for the public of data-hungry AI.
Reflecting on how AI affects society
As the public understands the potential of the technology, owners of AI models are facing government regulations, criminal suits and fears of potential security risks. This has caused great interest in the ethical, social, legal and safety aspects of this technology. “There has been an increasing number of workshops, researchers, research topics, groups and labs that are thinking about the social impact of AI, how to reduce the negative impact and emphasize the positive impact unstable,” said Weiwei Pan, an assistant. director of Graduate Studies in Data Science at Harvard University, who attends many AI conferences every year.
As the data used to train all these models is increasing, another area of research has become data – its sources, legal, regulatory and ethical issues regarding its purchase. One of the best paper awards went to ‘Measure Dataset Diversity, Don’t Just Claim It’, a research paper that focused on data diversity, and brought rules- a social science foundation for describing the diversity of databases.
“We need clearer definitions of concepts like diversity, which can mean different things to different people,” said Dora Zhao, a researcher at Stanford University and lead author of the paper. Like Zhao, many researchers pursuing PhDs are focusing their work on the social benefits of AI, which was not the case a few years ago.
Another paper that won the best paper award, debated two AI models to find the right answer for humans. “It makes us, the users, judge two AI experts arguing,” said Akbir Khan, the paper’s lead author, who is doing a PhD at the University of London. Khan who started out before going into research feels that with the many types of AI out there today, it should be easy for people to find the right answer between them all. His research is a first step in that direction.
Is AI only for the rich?
Science is never perfect, and that’s especially true for machine learning. Of the 10,000 people who attended this meeting, half were from America and Western Europe, followed by China which had about 1,000 people who attended the meeting. Conference expenses aside, there was little representation from Global South countries with only 100 from India. This is due to the high cost of doing AI research.
“The Global South lacks the computing, infrastructure and capabilities to compete in the AI space,” said Vukosi Marivate, a professor at the University of Pretoria in South Africa, co-founder of Africa’s largest Deep Learning Indaba conference. who gave the keynote speech at the meeting. “There is the US, China, Europe a bit, and there is also India which like everyone including Africa has scraps,” Marivate suggested that instead of chasing profits, countries these should focus on the needs of their community.
“I’m worried about whether companies are in charge of AI,” said Sarawagi, adding that computer science researchers should find the needs of their communities when it comes to this technology. “AI may help developed countries reduce their dependence on humans, but for us, it will not solve social problems such as living conditions, hunger or climate change,” he said. Instead of following the developed world, we should create our own systems, encourage local and relevant R&D, and change government policies.
Shweta Taneja is an author and journalist based in the Bay Area. His bi-weekly column will showcase how emerging technologies and technologies are changing society in Silicon Valley and beyond. Find her online with @shwetawrites. Opinions expressed are personal.
#Silicon #Social #anxiety #high #among #researchers