news

is ai an intelligent body that is both emotionally and intellectually superior, specifically designed to cure human emo?

2024-09-26

한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina

author | yugu

disclaimer | the title image comes from the internet. this is an original article from the jingzhe research institute. if you wish to reprint, please leave a message to apply for whitelisting.

as the first year of the big model implementation in 2024, people's exploration of ai has moved from the improvement of algorithms to the expansion of application scenarios. one of the most representative phenomena is that almost all big model companies and ecological enterprises have turned their attention to intelligent agents.

jingzhe research institute noted that at the 2nd baidu "wenxin cup" entrepreneurship competition, which just concluded at the 2024 baidu cloud intelligence conference, the number of participating projects has reached nearly 1,600, and the number of participants has almost doubled. among them are not only international professional teams, but also young forces from 300 universities.

in his video speech, baidu founder, chairman and ceo robin li specifically mentioned that in this competition, "more than 60% of the participating teams focused on intelligent applications, and more than 30% of the participating teams did not have professional programmers." this means that ai startups can already "go into battle lightly" by connecting to large models and using application development tools provided by platforms such as baidu to develop industry applications.

as the entire ai industry moves from the "battle of a hundred models" to the "competition for supremacy among intelligent entities" stage, the value of ai at the commercial level is also being continuously verified.

the large model lands and rolls towards the intelligent body

from conversational service agents to workflow orchestration agents, as an extension of big model capabilities, agents are being applied to various fields in order to find precise positioning for concrete products and services in complex and changing market demands.

according to the observation of jingzhe research institute, compared with the participating teams of the previous "wenxin cup",the biggest change this year is that the proportion of application layer projects in the participating projects has increased significantly, from 80% last year to over 90% this year, and the application directions are becoming more diverse.last year, more than 30% of the participating projects were concentrated in the general office and marketing fields. this year, they cover multiple fields including entertainment, e-commerce, marketing, medical care, office, hardware, and enterprise services.

for example, jirui technology, which won the first prize of this year's "wenxin cup", developed a one-stop e-commerce material ai tool to meet the diverse needs of e-commerce scenarios where products are updated quickly and a large number of different content materials such as pictures, texts, short videos, etc. are required.through the research and development and application of technologies such as computer vision, deep learning, image generation and editing, this tool can provide consumer brands with ai content generation, management tools and conversion services covering pictures, texts and short videos.while helping companies reduce product photography costs, it also improves the conversion rate of content marketing.

in the social field where ai is expected to have great potential, the winning team of this competition, kotoko, has created a multi-agent driven virtual character social game interaction platform.the uniqueness of this gaming platform lies in that, in addition to allowing users to create personalized ai characters, the underlying intelligent agent architecture enables intelligent interaction between "environment and character" and "character and character".it brings users a more realistic and fun-filled experience, which is somewhat like the ai ​​version of "the sims".

the "guided ai tutor" created by the "teacher ai" team based on the general education model abandons the traditional tutoring method of "giving answers directly".by providing interesting and personalized knowledge point summaries, guided questions and answers, encouraging answers and other functions, it helps students master knowledge points in the process of answering questions, uses ai to increase students' interest in learning, and helps improve their academic performance.

in addition, professional psychological diagnosis and treatment ai agents based on large models and social media marketing agents that assist enterprises in private domain traffic operations and improve customer acquisition and operational efficiency are also quite eye-catching.

in particular, the psychological diagnosis and treatment ai agent from mirror technology has chosen the "niche" track of psychological diagnosis and treatment in addition to the already highly internet-based industries such as design, social networking, and education.it not only challenges the possibility of ai's transformation from technical principles to market applications, but also explores the deeper value of ai to humans in the form of intelligent entities.

the big market behind “little emo”

in recent years, whether on social media or in real life, topics related to mental health such as emo and depression are constantly being brought up, and the "healing economy" is popular.

however, behind the growing demand, the psychological treatment market still has the industry pain point of insufficient supply.the blue book "china's mental health in 2023" points out that the current clinical diagnosis and treatment of mental and psychological diseases in my country faces huge challenges due to the high prevalence and large number of patients, but the domestic rate of depression treatment is only 9.5%.

this is because psychological diagnosis and treatment, as an uncommon medical need, has faced problems of shortage of professional talent and lack of medical resources in the medical supply system based on public hospitals in the past.

according to data from the national health and family planning commission, in 2020, there were only 40,490 practicing physicians in china's psychiatric hospitals, and 45,432 practicing assistant physicians, with a total of only 6.1 psychiatrists per 100,000 people, a significant gap compared to 12.7 in the united states and 11.9 in japan.

in addition, psychological diagnosis and treatment also incurs high costs in terms of time and money because it is highly dependent on one-on-one face-to-face consultations.according to the legal daily, because psychological counseling treatment takes a relatively long time to take effect, patients often need to go through five or six sessions, or as many as dozens of sessions, in one course of treatment, with costs ranging from several thousand yuan to tens of thousands of yuan.

precisely because the overall cost is too high, few patients in real life can persist in treatment until they are completely cured. therefore, whether for potential people or patients who really need it, psychological diagnosis and treatment has a higher treatment threshold than general diseases.

the "high threshold for medical treatment" is actually an efficiency issue. due to the constraints of the number of doctors and physical space, the "one-to-one" diagnosis and treatment capabilities cannot be replicated on a large scale. however, with the help of intelligent bodies, the diagnosis and treatment channels built online can break through the limitations of time and physics and meet the growing market demand.

for example, mirror technology's ai assessor uses adaptive scales and digital human video interview screening to help users quickly diagnose their mental health status. the "ai doctor" built using multimodal ai technology can also conduct natural interviews with users and provide evaluation suggestions.the most important thing is that users do not need to take leave or go to offline medical treatment. they can enjoy convenient medical treatment 24 hours a day, 7 days a week through online channels.

in addition, some people with sub-healthy mental conditions, such as mild anxiety, have not yet reached the stage where formal treatment is needed. at this time, they need companionship and someone to talk to. mirror image technology's ai confidant can provide them with light consulting services. whether it is a simple chat or "ai tree hole"-style listening, it can help users relieve stress and maintain mental health.

for mild to moderate patients, mirror technology's ai psychologists can assess the mental health of users based on their expressions based on natural language processing and emotion recognition technology, and then give positive feedback in a close-to-real way, providing personalized psychological counseling and treatment plans to help users relieve anxiety and resist depression.

to a certain extent, ai agents built around scenarios such as diagnosis, companionship, and treatment already have the capabilities of "ai doctors."in addition to leveraging the efficiency advantages of ai to meet personalized market demands in a "one-to-many" format, online diagnosis and treatment methods also play a role in protecting the privacy of minority groups such as minors.

in fact, the application of ai in the field of psychological diagnosis and treatment has always been controversial, because the main way to cure mental illness is still through long-term communication between psychologists and patients, and ai has always been considered to lack empathy and cannot accurately understand emotions. therefore, the effectiveness of the interview between "ai doctors" and patients is difficult to guarantee.

however, in mirror technology's three products, ai assessor, ai confidant and ai psychologist, we can see that ai can provide patients with an outlet for negative emotions and relieve their psychological pressure through "human-like" communication without the need for empathy.

in addition, although ai cannot understand emotions, based on the construction of professional theoretical knowledge models and natural semantic analysis, ai agents can also capture the causes of psychological stress during communication, thereby providing personalized solutions based on professional knowledge.

for example, when a user talks about his family during a conversation, the intelligent agent can guide the user from the perspective of family affection.in other words, ai may not be able to empathize, but it can conduct reasonable analysis and reasoning based on the questions currently raised by users, and then provide specific solutions based on professional knowledge.therefore, the role played by ai agents in the field of psychological diagnosis and treatment is similar to that of traditional treatment methods.

ai agent, the next level of "psychologist"

in fact, the application of ai technology in the field of psychological diagnosis and treatment is nothing new.

in the 1960s, joseph weizenbaum, a computer scientist at mit, developed a computer program called "eliza" to simulate "person-centered therapy." before chatgpt set off a new round of ai craze, there were also digital diagnosis and treatment methods developed based on intelligent voice interaction functions.

however, when compared with ai agents, it can be found that early ai applications in the field of psychological diagnosis and treatment are almost inseparable from the product form of chatbots, and have not reached the level of "ai doctors".this is because in the early stages, ai’s strengths are still focused on its powerful computing power, so more innovation is concentrated on developing psychological testing and assessment tools.

at this time, ai is like a machine that gives patients preset questions and collects answers, assisting psychologists and counselors in assessment and diagnosis. ai itself has no analytical ability, nor can it combine answers beyond the preset ones to make reasonable logical inferences.

therefore, although ai has been used in the field of psychological diagnosis and treatment for half a century, it has not yet had a decisive impact on the current status of the psychological diagnosis and treatment market.until the stunning success of chatgpt, more and more people have seen the comprehensive capabilities of large models in information collection, analysis, and feedback, and new ideas have been brought to the application of ai in the field of psychological diagnosis and treatment.

today's ai agents have long gone beyond the scope of "products".

compared with chatbots with single functions, ai agents that can automatically diagnose, provide 24/7 companionship, and provide personalized treatment plans not only have super computing capabilities, but can also autonomously perceive the environment, make decisions and perform actions. at the same time, they can adapt to different groups of people and different scenarios to provide a "human-like" interactive experience.

and as the large model grows, the intelligent agent can continue to evolve more comprehensive capabilities.for example, in the marriage and love scenarios other than psychological diagnosis and treatment, mirror image technology has launched a variety of ai agents such as ai matchmaker, love assistant, and marriage relationship counselor. with the comprehensive capabilities of ai agents and the diversification of application scenarios, the psychological diagnosis and treatment industry will also enter the "next level".

it should be pointed out thatthe evolution of psychological diagnosis and treatment ai agents is actually a process from meeting public needs to solving professional problems, which is inseparable from the feeding of effective information and the supply of professional capabilities.

unlike simple mathematical calculations, the reasoning logic of psychological diagnosis and treatment is based on theoretical knowledge and real cases. ai agents need more accurate interfaces to obtain massive amounts of valid data, from which they can distinguish and accumulate enough "user interview cases" to build their own analysis models, so that they can accurately find the user's problems and match the corresponding solutions in subsequent diagnosis and treatment.

when giving feedback to users, ai agents also need to choose appropriate ways of expression based on the other person's personality traits and different scenarios. for example, they can be more sincere when accompanying others, or turn into a joke teller to occasionally make a joke, or be more authoritative and professional when providing treatment plans. this requires the use of nlp or the ability of dialogue and cross-language models.

therefore, returning to the industrial end, whether ai agents can reflect their value in application scenarios depends largely on the capabilities of the big model it adopts, and baidu wenxin's big model provides fertile soil for the growth of ai native applications.

at present, wenxin big model includes basic big models such as nlp, cv, and cross-modality, task big models such as dialogue, cross-language, search, and information extraction, big models in the field of biological computing, industry big models, and tool platforms to support the application of big models, forming a three-level big model technology system of foundation-task-industry, with two major characteristics of knowledge enhancement and industrial level, which is sufficient to meet the development needs of intelligent entities in different fields.

in addition to the comprehensive capabilities of the wenxin big model, baidu's ecosystem's empowerment of intelligent entities is also quite valuable.

for example, baidu smart cloud qianfan appbuilder, as an industrial-grade ai native application development platform, can provide services in both code and zero-code forms, helping developers to continuously lower the threshold for application development. baidu search has currently become the largest distribution portal for intelligent entities, distributing more than 10 million per day, helping more intelligent entities seize market opportunities.

in addition, in may this year, baidu made three lightweight models, ernie speed, ernie lite and ernie tiny, available for free, and in july, it significantly reduced the prices of two flagship models, wenxin large model 3.5 and 4.0. for intelligent body startup teams that are still exploring "technology monetization", this is also a more practical growth condition.

huang li, founder of mirror image technology, also said in an interview with jingzhe research institute that "wen xin yi yan has greatly improved roi."

summarize

although ai has attracted much attention, many people still believe that it is difficult to implement ai applications.

surprisingly, in the application of psychological diagnosis and treatment intelligent agents,the implementation of ai technology not only solves the pain points of the industry and lowers the threshold for ordinary people to access psychological diagnosis and treatment, but also to a certain extent reflects the value of ai accessibility and accomplishes a "difficult but right thing."

in fact, baidu has been holding the baidu "wenxin cup" entrepreneurship competition since last year, and has invested tens of millions of yuan in participating projects and provided various ecological support. baidu is also doing the "difficult but right thing."under the new trend of shifting from large models to intelligent entities, the difficulty of ai entrepreneurship has actually been greatly reduced.

first of all, from the “evolution” from large models to intelligent agents, we can see that with the continuous iteration of technology, the usability of ai is also constantly improving, and ai startups in the form of intelligent agents no longer require “hand-coding”.

in the words of robin li, "intelligent entities are equivalent to websites in the pc era and accounts in the self-media era. their most obvious feature is that the threshold is low enough that anyone can get started, and the ceiling is high enough that very complex and powerful applications can be made." based on baidu's wenxin big model and qianfan intelligent platform, current ai technology can already support teams and even individuals to build intelligent entity applications with user value and commercial value.

secondly, the winning projects of the "wenxin cup" such as the psychological diagnosis and treatment intelligent agent also prove that valuable intelligent agents have found the intersection of ai technology and commercial value, and it may only be a matter of time before intelligent agents in different tracks can achieve precise matching of products and market demands.

third, as more intelligent agents emerge, each field will generate unique intelligent agents based on its own specific scenarios and needs in the future, forming a huge ecosystem of millions of agents. at the same time, the capabilities of intelligent agents are also constantly improving. in the future, intelligent agents may also have the ability to collaborate and complete more complex tasks or needs.

by then, the ai ​​agents with both emotional and intelligence excellence will be able to solve more than just human emo problems.