2024-05-26 10:42:05
[위클리 리포트] AI serving to the socially underprivileged
Simultaneous processing of audio-visual info ‘multimodal’ expertise acts as eyes and ears
MS developed an artwork commentary service, and OpenAI developed a voice assistant.
Samsung releases dwelling home equipment with care features… Aged care robots are additionally evolving
《‘Good AI’ amid ethics-safety issues
As issues concerning the misuse of synthetic intelligence (AI) expertise develop, international locations around the globe are introducing laws on AI ethics and security. Nevertheless, then again, providers are being launched to assist socially deprived folks such because the aged and the disabled. AI explains issues to the visually impaired as in the event that they have been drawing a murals, and checks on the well-being of aged individuals who dwell alone. We glance into the world of ‘good AI’, which is being reborn because the eyes and ears of the disabled and a companion for the aged.》
“’Self-Portrait’ is a portray painted by Vincent van Gogh in 1887 utilizing cardboard and oil paint. The principle character of the portray is a person with a purple mustache and beard that’s barely brown and white. His inexperienced eyes stare instantly on the viewer, his pores and skin is pale and his cheeks are pink.”
This content material, which appears to be defined by a curator in entrance of a murals, was written by Microsoft’s synthetic intelligence (AI) service ‘Co-Pilot’. Microsoft just lately started working with the Rijksmuseum in Amsterdam to create detailed explanations for about 1 million artistic endeavors utilizing AI. That is to supply explanations of artistic endeavors to visually impaired individuals who can not totally take pleasure in artistic endeavors.
An official from the Amsterdam Museum stated, “If it weren’t for Co-Pilot, what would have taken years might now be accomplished in only a few hours.” Karin de Brawin, who suffers from Usher syndrome, which causes her listening to loss and blindness, stated, “The detailed descriptions generated by AI enable me to image the work in my head.”
There are lots of issues that AI expertise will produce deepfakes and false info and pose a menace to human life. Voices calling for laws for security and ethics are rising. Nevertheless, because the AI competitors triggered by ChatGPT is mixed with numerous info and communication applied sciences (ICT), the variety of instances of optimistic features getting used to assist susceptible teams such because the disabled and the aged are rising.
● Massive tech’s ‘multimodal’ expertise serves because the eyes and ears of the disabled
Since final yr, huge tech corporations have been growing ‘multimodal AI’ expertise that goes past paperwork and processes photos, voice, and textual content concurrently. Consultant examples embody OpenAI’s ‘GPT-4’ and Google’s ‘Geminii’. As AI’s voice and picture processing capabilities have change into extra refined, it has change into attainable to play the function of individuals’s eyes and ears.
The visible and voice processing capabilities of ‘GPT-4o’, a brand new voice assistant service launched by OpenAI on the thirteenth (native time), are anticipated to be useful to the visually impaired and others. GPT-4o encompasses a quick response velocity of about 0.32 seconds, which is greater than 10 instances improved in comparison with its predecessor. Which means rapid human-level assist may be offered to the disabled.
On the identical day, OpenAI launched a video of a visually impaired person utilizing GPT-4o. When a person requested in entrance of Buckingham Palace, “Is the king staying within the palace proper now?” GPT-4o responded, “I feel it’s as a result of I see the royal flag, which suggests the king is staying.” Once I informed him I wanted to take a taxi once more, he instantly stated, “There’s a automobile coming in direction of you on the left aspect of the street. He even provided assist by saying, “Get able to catch a taxi.”
Google, which held its annual developer convention I/O a day afterward the 14th, additionally emphasised the social inclusiveness of expertise. On today, Google introduced that the multimodal expertise of ‘Gemini Nano’, a small language mannequin (sLLM) within the type of ‘on-device AI’ that operates with out being linked to the Web, can be mixed with ‘Talkback’ within the second half of the yr (July to December).
Talkback, developed for the visually impaired, is a perform that reads textual content on the display aloud. When mixed with multimodal expertise that may extremely course of photos and voices, it’s anticipated that clearer and richer info can be delivered to the visually impaired.
For instance, when buying a costume on-line, Gemini acknowledges the form of the clothes on the display and offers voice steering to the person, resembling “This costume is brief, has a collar, lengthy sleeves, and a big ribbon on the waist.” It is going to be delivered to assist customers make purchases. As well as, Google additionally helps the ‘Lookout’ perform, which makes use of the cell phone digital camera to elucidate detailed details about the environment.
● Home massive companies introduce specialised providers for the aged
Home corporations have additionally just lately been introducing info expertise (IT) resembling AI that may assist the aged and the disabled. On the 14th, Samsung Electronics unveiled its ‘Household Care’ service that helps household care by means of AI. Household Care is a service that applies AI to on a regular basis dwelling home equipment and sends notifications to youngsters when irregular life patterns are detected.
For instance, if a mum or dad doesn’t take treatment, the speaker publicizes the preset time to take the treatment by voice, and when the drawer containing the treatment is opened, the treatment document is saved. Moreover, if the house equipment isn’t used for a preset time frame by the guardian, it detects this as an abnormality and offers a notification. Samsung Electronics introduced that it will possibly test whether or not mother and father are utilizing TV, fridge, water air purifier, induction, and so on. by means of ‘Sensible Issues’, an built-in connection platform, and that it’s also attainable for youngsters to manage dwelling home equipment remotely.
Round October, Samsung Electronics plans to introduce a perform that makes use of the digital camera put in contained in the robotic vacuum cleaner to detect an individual who has collapsed in the home and supply a notification service.
Naver, which owns a large-scale AI ‘Hyperclova’, is utilizing Hyperclova to supply ‘Clova Care Name’ to single senior residents who want care. This can be a service that mechanically calls the aged and asks how they’re doing on subjects resembling weight loss plan, sleep, and well being. Naver stated that Clova Care Name not solely offers “sure or no” solutions, but in addition permits pure dialog, which may be anticipated to have the impact of assuaging loneliness and isolation among the many aged.
Clova Care Name started a pilot service in Haeundae-gu, Busan in 2021 and was formally launched in Might 2022. At the moment, it has been launched in about 110 cities, counties, and districts and is offering providers to about 24,000 folks.
● Startup that developed providers for the disabled
Not solely massive companies, but in addition startups and social ventures are providing IT providers for the socially deprived. A consultant instance is TwoArt, which launched ‘Sullivan Plus’, an help utility (app) for the visually impaired in 2018. Sullivan Plus is a service that mixes smartphone cameras and AI expertise to acknowledge objects in entrance and transmits this by means of voice to assist visually impaired folks transfer or acknowledge objects.
Once I ran this app, I noticed, “There’s a laptop computer on the desk and books on the bookshelf. He described the scene in entrance of him intimately, saying, “The door is closed and there’s a title tag on the door.” Utilizing the textual content recognition perform, even the title of a small e book was acknowledged and offered as a voice. The service is at present being utilized by roughly 365,000 visually impaired folks in additional than 200 international locations. In January of this yr, the corporate additionally launched a brand new app known as ‘Sullivan Finder’, which is a complicated app that provides features resembling search.
TwoArt, which has been collaborating with SK Telecom since 2021, just lately put in SK Telecom’s multimodal AI ‘Adot X’ on Sullivan Plus and Sullivan Finder. The AI can study greater than 1 billion photos and describe photos extra exactly. Soo-won Cho, CEO of ToArt, stated, “After the brand new multi-modal AI was utilized, the extent of expertise has improved and the variety of customers has additionally elevated.”
Hyodol, a home firm, has been distributing ‘Hyodol’, a pet robotic geared up with AI expertise, to native governments throughout the nation since 2018. Roughly 10,000 aged folks in 156 native governments throughout the nation are utilizing robots. Once you contact its head or hand, it acts cute like a seven-year-old grandchild or offers reminders to eat, train, take drugs, and so on. at set instances. When an aged particular person asks Hyodol for assist in an emergency scenario or when no motion is detected at a sure time, an emergency connection is made to the emergency management middle or guardian.
The corporate emphasised Hyodol’s ‘emotional connection perform.’ Because of Hyodol distributing robots to 186 aged folks residing alone within the Jeollanam-do area from 2021 to 2023, the outcomes confirmed that their psychological well being standing improved, together with a lower within the aged melancholy scale from 10.29 to 9.05 and a suicide ideation scale from 15.52 to 9.25. He stated he noticed it. Based mostly on these achievements, South Jeolla Province is investing a complete funds of three billion received from 2023 to 2025 to introduce 1,100 Hyodol models to all 22 cities and counties.
Final yr, the corporate launched ‘Hyodol 2.0’, which is an upgraded model of the earlier model. In Hyodol 2.0, ChatGPT was launched and the communication perform with the aged was strengthened. A Hyodol official defined, “The attribute of Hyodol 2.0 is that extra pure ‘two-way conversations’ are attainable.”
One other home firm, Sound Viewing Path, offers real-time subtitle providers for the listening to impaired. AI acknowledges sounds within the discipline in actual time and shows subtitles as textual content. Based on the corporate, real-time subtitles confirmed a excessive accuracy of 96-97%. This firm has been offering providers to over 800 establishments, together with faculties and banks, since 2017, and has expanded its providers to basic customers beginning this yr.
● Inclusive service acknowledged by the worldwide group… The federal government additionally rolls up its sleeves
TwoArt and Hyodol, together with massive corporations resembling Samsung Electronics and SK Telecom, additionally received the ‘World Cell Award (Glomo Award)’ on the ‘Cell World Congress (MWC) 2024’ held in Barcelona, Spain in February. The Glomo Award, chosen by the World Cell Communications Affiliation (GSMA), is an award so prestigious within the ICT business that it’s known as the ‘Oscars of the cell business.’ TwoArt’s Sullivan Finder received the ‘Greatest Cell Use Case for Accessibility and Inclusiveness’ class, and Hyodol’s Hyodol Robotic received the ‘Greatest Innovation for Well being and Properly-being’ class.
The federal government and native governments have additionally rolled up their sleeves to supply social inclusion providers by means of IT. The Ministry of Science and ICT and native governments have been finishing up the ‘Sensible Village’ undertaking since 2019 to enhance inhabitants decline, growing old, and resident welfare. The undertaking, which began with 4 billion received in 2019, was allotted a funds of 103.9 billion received this yr.
Particularly, the ‘Sensible Senior Middle’ undertaking, which Bucheon Metropolis, Gyeonggi Province, started working for the welfare of the aged, has been acknowledged for its achievements and has now expanded to 27 native governments. The good senior middle offers leisure and well being packages by means of a non-face-to-face video system and healthcare packages by means of Web of Issues (IoT) units. By this, aerobics, laughter remedy, and well being counseling may be carried out remotely, and blood stress and blood sugar ranges may be measured periodically. An official from Bucheon Metropolis stated, “We plan to extend the variety of good senior facilities at present put in within the metropolis from 45 to 150 by 2026.”
Reporter Jeon Nam-hyuk ahead@donga.com
2024-05-26 10:42:05