the number of participants at the bund summit hit a new high, and people are paying attention to how to prevent ai from doing evil
2024-09-08
한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina
on september 7, the 2024 inclusion·bund conference officially closed. in fact, on that afternoon, a few hours before the conference closed, there were still audiences queuing to enter. official data showed that this conference attracted a total of 52,000 on-site participants, setting a record high in terms of both the scale of participants and the number of international guests. in addition to young people, the reporter noticed that there were also many elderly people and children in the audience.
the theme of this year's bund conference is "technology creates a sustainable future", including 36 open insight forums, more than 10 innovator stages, 2 ai innovation competitions, a 10,000 square meter exhibition, a green life festival and a technology talent recruitment fair. according to statistics, more than 30 new ai achievements were released at this year's bund conference within three days, including ant ai life manager "zhi xiaobao", ai financial manager "ant xiaocai", ai health manager, etc.
the 10,000-square-meter exhibition venue brought together many companies to participate in the exhibition. the reporter observed that compared with previous years, more and more manufacturers have prepared various equipment experiences and interactions. in front of many booths, the audience lined up to participate in the experience. paying attention to humanity has become a major feature of this bund conference. in addition to the touching moment at the opening ceremony when the 24-year-old one-armed boy zhou jian played the piano with his intelligent bionic hand, in the exhibition area of the bund conference, with the help of equipment, the audience can experience the visualization of brain waves in the form of flowers, so that the momentary emotions can be commemorated. there is also a device that can realize "hug transmission". wearing a special vest with embedded artificial muscle fibers, when another person hugs a model equipped with a pressure sensor, the experiencer can feel the "remote hug" that simulates its voice control, tactile feedback, temperature regulation and other sensory elements.
at this year's bund conference, discussions and research on how to use "technology for good" received more attention from the industry.
ai medical care attracts attention
during the three-day exhibition, ai medical related products and corporate booths always had the largest flow of people. "what should i do if my parents' memory declines?" "how can ai assist in cancer screening?" these questions made many visitors unconsciously step into the relevant exhibition areas. more and more elderly people joined in brain health testing, and young people joined in the "drawing" of brain waves.
"compared with last year, the share of ai in this year's bund conference seems to be higher, and it's not like the silly ones before. many applications this year are really useful. it is obvious that the algorithms have been improved a lot. for example, facial recognition function or voice interaction function feels like it has been upgraded to a higher level, and the effect has become better, so everyone is enthusiastic about experiencing it." an exhibitor at the conference told reporters.
before i had time to take a sip of water, the staff at the damo academy ai medical booth welcomed the next visitor for consultation. at this year's bund conference, the damo academy medical ai laboratory exhibited its ai medical imaging analysis product. with the help of ai technology, this product can screen for eight major cancers, including pancreatic cancer, esophageal cancer, intestinal cancer, gastric cancer, and a series of chronic diseases from plain scan ct scans, attracting many visitors.
the staff at the booth told reporters that through ai identification, this product can achieve an accuracy rate of more than 90% in screening for cancers such as liver cancer and lung cancer, improve doctors' work efficiency, and make up for some problems caused by the imbalance of medical resources.
"in fact, this technology is mainly for the b-end, serving doctors, and its clients are mostly hospitals, experience agencies, insurance companies, non-profit organizations, and software manufacturers, but many individuals are also very interested in it these past two days." the above staff member said. although many ai medical products are not directly for c-end users, the audience at the bund conference is curious about how ai medical care will change the diagnosis and treatment of diseases.
the booth of bosten technology attracted a large number of elderly visitors, many of whom took out their mobile phones and answered questions through alipay's "touch and pay" function to test their risk index for alzheimer's disease. according to the staff of the exhibition area, with the support of many data and materials, this product can detect alzheimer's disease through answering questions, voice interaction, motion recognition, etc., and generate a user-specific brain health training plan through a large model.
"our product was completed in 2020, but the dynamic recognition function with the help of the camera was just launched this year, mainly due to the development of ai functions, including the recommendation of training plans, which is indeed much smarter after the appearance of the large model." the staff of the exhibition area told reporters. with the development of ai, the "evolution" of more ai medical products is worth looking forward to.
it should be emphasized that, at present, the diagnosis and treatment effects of many ai medical products still lack public data or verification by third-party authoritative organizations. the staff at the relevant booths also said that ai medical products cannot completely replace the diagnosis and treatment needs of doctors, and mainly play an auxiliary role. however, the application of ai in medical problems is indeed expected by the industry.
many large companies are also betting on the ai medical track. at the bund conference, alipay released a new ai application product "ai health manager", which provides users with more than 30 ai health services such as finding doctors, reading reports, accompanying doctors to see doctors, asking about medical insurance, and managing health. wang xiaochuan, founder of baichuan intelligence, introduced that baichuan intelligence will focus on investing in the field of large-scale medical ai, starting from the two entry points of pediatricians and general practitioners, hoping to achieve low cost and accessibility of medical care next year and reduce the phenomenon of hospital runs.
zhou hanmin, member of the standing committee of the national committee of the chinese people's political consultative conference, said at a forum at the bund conference that with the continuous development and application of new technologies such as artificial intelligence, the medical industry is ushering in a profound change. according to his observation, from the initial imaging diagnosis, auxiliary diagnosis and treatment, and genetic testing to drug discovery, clinical trials, and health management in recent years, artificial intelligence has significant advantages in improving accuracy and efficiency, personalized treatment, and data processing. he pointed out that the integration of medical health and artificial intelligence has also extended to the field of traditional chinese medicine.
talking about the application of artificial intelligence in the medical industry, zhou hanmin said that the application of artificial intelligence in the medical and health industry should focus on improving accuracy, applicability and safety. in addition, attention should be paid to whether the introduction of artificial intelligence will lead to atrophy of medical staff's skills, misjudgment of trust in artificial intelligence, and information overload.
fighting ai with ai
while we expect technology to bring us a better life, how to prevent technology from doing evil has also become a hot topic of discussion at the bund summit.
at this year's bund conference, an ai innovation competition was held, including the "afac2024 financial intelligence innovation competition" and the "global deepfake attack and defense challenge". due to the recent "deepfake" incident in south korea, which is believed to be a reappearance of the "nth room" case using new criminal methods, using ai against ai, the deepfake attack and defense challenge was particularly challenged by the audience.
in fact, with the development of ai technology, deepfake has spawned a black industry chain such as fraud. the economic and property losses, reputation damage and other incidents caused by ai face-changing fraud are becoming a global challenge.
at the bund conference exhibition area, many viewers took photos on the spot, and then used the deepfake face-changing model and algorithm to replace the audience's real photos through video activation, smile and glasses templates, and personally experienced the realism of deepfake.
at the bund ai innovation competition global deepfake attack and defense challenge, the audience was even more excited about ai identifying aigc's fake pictures. in this show, the host took out 9 photos of actresses such as marilyn monroe, 2 of which were forged by aigc. the contestants used their own models to give the probability of forgery for each picture, and compared the 2 photos with the highest probability values with the correct answer. in the end, the models of both teams found the correct answer.
in the subsequent interaction, an interesting phenomenon is that ai may not be able to identify fake images that are "fake at first glance" that can be seen by the human eye, but ai can identify some images that are difficult to distinguish between true and false with the naked eye. professor zhou tianyi of the agency for science, technology and research of singapore analyzed that humans rely on intuition and logic to judge whether an image is fake. some problems that humans can see at a glance cannot be identified by ai in terms of relevant background and situation. however, for some difficult-to-identify images, ai's meticulous and diligent work mode of completing human instructions can actually find problems. he said that compared with "fake at first glance" images, it is precisely these difficult-to-identify images that often cause greater losses to humans. therefore, it is of great significance to use ai to deal with deepfake.
"from the data we are seeing now, the volume of attacks, whether we call deepfake or aigc-generated forgeries, has increased about 10 times this year compared to last year. last year, the concentration of attacks was about 1%, and this year it is 10%. the entire risk is actually rising exponentially. this is not only happening in china, but also in south korea, the philippines, indonesia and other countries." yao weibin, the creator of the bund conference ai innovation competition and global deepfake attack and defense challenge and technical director of zoloz, said that the 10% concentration means that the industry has discovered that 10% of online traffic is deepfake traffic. with the development of aigc, the cost of generation will become lower and lower, and the efficiency will become higher and higher. next year, it may not be 20%, but may be 50%. this is a threat that must be taken seriously.
more ideas on how to guide ai for good were also discussed at the bund conference. yao weibin admitted that developing detection algorithms is only one way to curb the abuse of deepfake technology. this method is relatively backward. if we want to prevent fraud from the root, yao weibin believes that we should require the establishment of a universal identifier for aigc content from the perspective of standards and legislation. for example, every video or picture generated by aigc has a signature or watermark in the file information, so that the platform management is more reliable and aigc can be more credible.
"let ai help people play the piano, not do evil." this is the value repeatedly emphasized at this year's bund conference. xiao yanghua, director of the shanghai key laboratory of data science, said that large-scale abuse of ai may damage people's nature as human beings. we should do what we should do and not do what we should not do. we should be cautious and restrict any ai application that harms people's nature as human beings. to this end, "we must attach importance to technology governance and consider governance in advance to prepare for a rainy day."
han bicheng, founder and ceo of zhejiang qiangnao technology, said that the existence of the ai ethics committee is very important. when technology advances, we should pay more attention to those who "move slowly". "when a technology that can change the world appears, we will use it to help those who need it most."
(this article comes from china business network)
