"Internet insurance" may be dead, but technology is immortal, and the AIGC track is booming|Recommended reading

Editor's note

While people are commemorating the decade of Internet insurance, the discussion about AIGC in the insurance industry is gradually escalating. Standing today, looking back on yesterday, "Internet insurance" gives people a feeling of being away from the world.

Under the full integration of online and offline, the boundary between the Internet and other channels has gradually blurred, and many businesses have become difficult to distinguish whether they belong to the real "Internet insurance". The business process that penetrates all products is almost inseparable from various technologies represented by the Internet.

Science and technology are developing too fast, penetrating too fast.

In 2013, the so-called "first year of Internet insurance", the Internet was still more of an emerging sales channel, and the research and development of scenario-based and fragmented insurance products had just begun, but in the following years, information technology represented by the Internet, mobile Internet, and even big data, cloud computing, and artificial intelligence gradually showed great influence on the insurance industry.

People's focus has gradually shifted from "Internet insurance" to "insurance technology", from the front end of the business to the whole process, iterating the system, upgrading the middle office, improving the process, innovating the model... The impact of technology on the insurance industry has deepened. A large number of different types of insurtech companies have risen in this process and have been favored by capital for a period of time.

Today, capital is ebbing, but technology and the insurance industry are so deeply bound that they can no longer be separated, and it can even be said that to a large extent, technology is reshaping people's imagination of the future of the insurance industry.

Especially at the end of 2022, the beginning of 2023, ChatGPT became popular, and then the release of GPT-4 in mid-March, people are full of expectations for the capabilities of large models, and some people even think that the dawn of the AGI (General Artificial Intelligence) era has appeared.

At present, the AIGC boom (AI-generated content) is intensifying: professionals praise, a large amount of money pouring in, a large number of startups are born, giants are in full power, and new actions of large models are frequent. What is even more exaggerated is that these extremely rapid changes have made many senior AI experts shout "can't keep up with the trend".

The term "Internet insurance" may no longer be able to describe the current situation, but the transformation of the insurance industry by technology has touched the soul and is bound to continue for a long time.

So, in the upcoming new era of AGI, what do we need to understand and pay attention to, how should we view the impact of AIGC development on all walks of life, and what opportunities and challenges will arise? To this end, WiseWin has selected several discussion articles on AIGC and the Big Model, trying to show them from multiple angles to help readers still observe and think soberly in the drastic world, not to be left behind by the times, let alone be caught up in the torrent.

Here are excerpts from the 10 articles and their contents:

What exactly are we talking about when we talk about AIGC

Read what AIGC, ChatGPT, and large models are in one article (simple version)

Latest roundup: What exactly is AIGC? What are the applications? (Roundup Version)

AI-generated content refers to blogs, marketing materials, articles, and product descriptions created by machines, among others. In general, AIGC can be divided into text, image, and video generation.

AIGC consists of three key components: data, hardware, and algorithms. High-quality data such as audio, text, and images are the basic building blocks of training algorithms. The amount of data and the data source have a critical impact on the accuracy of the prediction. Hardware, especially computing power, makes up the infrastructure of AIGC.

AIGC has three cutting-edge capabilities: digital twins, intelligent editing, and intelligent creation. These functions are nested and combined with each other, giving AIGC superior spawning capabilities.

The AIGC industry chain is an interconnected ecosystem from upstream to downstream. Data vendors, algorithm agencies, and hardware development agencies are the main components of upstream AIGC. The midstream industry includes large technology companies that integrate upstream data, hardware, and algorithms. The downstream segment is mainly composed of various content creation platforms.

With the rapid development of hardware and algorithms, AIGC is expected to have more substantial applications in the future, and the most promising directions include cross-modal generation, search engine optimization, media production, e-commerce, film production and other fields.

AIGC in China

Lu Qi's latest speech transcript: My big model world view

In his speech, Lu Qi fully explained the macro thinking in the era of big models, the internal motivation of inflection points, technological evolution, structural opportunities for startups, and suggestions for entrepreneurs.

He believes that anything that changes society and industry is always a structural change. This structural change tends to be a large type of cost, moving from marginal cost to fixed cost.

And what is the inflection point in 2022-2023? It is unstoppable and unstoppable, and why? Same. The cost of the model goes from marginal to fixed because there is something called a big model.

Lu Qi said that the next inflection point will be the combination: "action" is everywhere (autonomous driving, robotics, spatial computing). First, in the next 15-20 years, models are knowledge and will be everywhere; Second, in the future, automated and autonomous actions can be everywhere; Third, people and digital technologies co-evolve, and co-evolution can achieve general intelligence (AGI). The four elements of general intelligence are: emergence + agency + affordence + embodiment.

AIGC in China

Chinese ChatGPT "Great Leap Forward"

Before and after Microsoft made ChatGPT a high-profile push to New Bing, Silicon Valley giants began to push for big model research, and Google released a ChatGPT-like Bard in just two months.

In this regard, China is not far behind. Since February 2023, Baidu, Alibaba, Tencent, JD.com, Byte, etc. have spoken out that they have carried out in-depth research in the field of large models and have obtained a lot of results. For a time, chasing large models has become a standard action in the domestic AI industry, and the transition period of "training models to refining large models" seems to be nearing the end, and the next stage has the posture of "big models for the whole people, ChatGPT into 10,000 homes".

Rush into AIGC in all walks of life

Considering that the domestic large model is not fully mature at this stage, it is more to learn from overseas paths to deduce domestic development. At present, overseas applications have blossomed, and a large number of product innovations have appeared in the fields of office, search, education, conversation and social networking, games, finance, e-commerce, pictures and videos, and related products and scenarios are expected to be further expanded after the launch of GPT plugin.

AIGC Feature: Large Models and AIGC Market Opportunities

Explore over 30 insurance scenarios! ZhongAn released the industry's first AIGC application white paper

Returning to the insurance industry, on May 23, ZhongAn Insurance and ZhongAn Technology jointly released the first AIGC application white paper in the domestic insurance industry, "AIGC/ChatGPT Insurance Industry Application White Paper".

Through expert research, the white paper sorts out more than 30 specific application links and scenario application points of AIGC technology in the insurance field, and forms a pre-judgment on the feasibility of technical implementation of application scenarios from multiple dimensions.

The natural language processing capability of the large language model (LLM) provides many possibilities for enterprise application landing scenarios, such as intelligent customer service, enterprise internal knowledge/rules and regulations, copywriting generation, and R&D operation and maintenance scenarios to improve efficiency.

Problems are starting to emerge

Behind the "AI Sun Yanzi" screen, how does AIGC face the pain of infringement?

Tencent Releases AIGC Development Trend Report: Embracing the Next Era of AI

Sun Yanzi is "pensioning", but AI Sun Yanzi has quietly become the traffic password of the entire network.

The technological changes triggered by ChatGPT have spread rapidly, directly impacting the field of content creation, and even evolving AIGC into an important part of the content platform. However, behind the national fanaticism, the development of AIGC also faces many challenges in science and technology governance issues. At present, there are mainly four challenges in the areas of IP, security, ethics and the environment.

First of all, the risk of new copyright infringement caused by AIGC has become an urgent issue facing the development of the entire industry. Due to copyright disputes, artists on ArtStation, a foreign art platform, have launched a boycott of AIGC-generated images.

Second, security issues always exist in the development and application of science and technology. In AIGC, it is mainly manifested in information content security, new illegal and criminal acts such as fraud caused by AIGC abuse, and the endogenous security of AIGC. The more famous case is that the fraud team used AIGC to change faces to fake Elon Musk's video, and defrauded digital currency worth more than 200 million yuan in half a year.

Third, ethical issues such as algorithmic discrimination remain. For example, DALL· E2 has significant racial and gender stereotypes.

Finally, there is the environmental impact, AIGC model training consumes a lot of computing power, and the carbon emissions are staggering. Previous studies have shown that the carbon emissions produced by the training of a single machine learning model are equivalent to 5 times the carbon emissions over the life of an ordinary car.

Other related reading

The latest interview with the CEO of OpenAI, 30,000 words detailing technology, competition, fear, and the future of humanity and AI

Lex Fridman, a research scientist at MIT and an AI researcher, spoke with Sam Altman, founder of Open AI. During the conversation, Sam Altman showed calmness and infinite optimism about the society brought by AI, which may be the motivation for his sincere enthusiasm to promote the continuous upgrading and iteration of GPT, while Lex Fridman expressed more concerns.

Sam Altman said: "I think there will be a lot of AGI in the world, so we don't have to compete with everyone. We will contribute some, and others will contribute some. It's good that these AGIs will be built differently, what they do, and what they focus on. ”

For what AGI means, Sam Altman argues, it is the culmination of amazing human endeavor. All the work, hundreds of thousands of millions of people, whatever it is, from the first transistor to packing the numbers we make into a chip and figuring out how to connect them together, and everything else, like the energy required. Science, like every step of the way, is like the output of all of us.

Microsoft CEO: The rapid development of AI is not terrible

After Musk and a group of scientists jointly issued an announcement calling for a moratorium on AI development, Microsoft CEO Satya Nadella was asked "Is AI developing too fast?" "AI didn't come out of nowhere, and search is an automated program, but we're moving toward a higher level of automation." AI is moving fast, but don't worry, it's moving in the right direction that humans can control. ”

He argues:

Although the 1.AI is developing rapidly, its development direction is correct, and it is developing rapidly in a direction that human beings can better control.

2. We all want to reap the benefits of AI technology and mitigate its adverse effects. However, to achieve this level, it is necessary to widely participate in all aspects of society, not just experts and scholars talking to themselves.

3.AI is actually a tool that gives people easy access to new technologies and knowledge, which improves the learning curve.

4. We are the only profitable company that is comfortable with the nonprofit control technology. I welcome others to do the same, and rightly so.

5. I think AI regulation is not about what will happen to the technology, but more about how to ensure that these technologies are deployed effectively to maximize their benefits and maximize their benefits.

6. It will all depend on the market suitability of the product, which is not a game for Microsoft and Alphabet. For me, things like OpenAI, Bing, or other small companies entering the search market and competing with us are all things to celebrate, and I'd love to see that.