#头条创作挑战赛#
Author / Baked buns under the stars
Editor/Spinach Starfield
Typesetting/sandwiches under the stars
The wave of artificial intelligence that began at the beginning of this year has driven many relevant players to take the "high-speed train". For example, in the field of AI hardware, NVIDIA's stock price and performance have achieved double growth, and for now, at least in the short term, this growth is sustainable.
In another area, AI software, is also considered by many investors to be on the eve of growth. The author recently saw that some positions in the market have clearly proposed that candidates should be proficient in AI application software, including ChatGPT. If there's certainty to be found in an uncertain market, perhaps it's certainty.
Since it is about to break out, many players are also planning ahead and actively preparing for fear of being left behind by the army. For example, creative software leader Adobe has launched a differentiated AI creation aid tool Firefly a few months ago, although it is still in the public beta stage, but has filled investors' expectations, and the stock price has risen by more than 50%. It is understood that the first model of the Firefly Model Set was trained on a unique dataset.
Gains in Adobe's stock price
In addition, some well-known brands have also "rubbed" the popularity of AI software, such as Xiaohongshu's company added AI software development business, YTO Express also established a technology company, the scope of business includes AI basic software research and development.
In fact, is the AI software track really that good? Is there an investment opportunity in this track? The author will take you to find out.
First, the dilemma
If you look back at the development of AI models in recent years, the number of AI technology providers is decreasing, and the threshold is constantly increasing. Especially in the era of large models in the past two years, the model has shown geometric growth, which has made the players with more data and financial support strong and strong, and the industry has shown a Matthew effect.
When it comes to AI models, in a nutshell, there are two ways to develop: open source and closed source. From the perspective of the route of the head players, OpenAI, which is popular on the whole network this year, has voted with its feet and chosen the closed-source mode. In this way, if other players want to integrate into GPT-4 in the future, OpenAI will provide the corresponding API interface. OpenAI is ready to zoom in two days ago, and it plans to launch a large model store similar to the "App store". Connect all ChatGPT software applications.
In the future, the author predicts that it is more likely to build an open source community and iterate models together to close the gap, but there is still a big gap with Google and OpenAI at the head.
The AI model changes gradually
The flame of collecting firewood is not necessarily high, and it has been vividly reflected in the AI model.
In this context, software players need to seriously consider how to stand in line, if you choose the head of closed-source players, although it can simplify training, deployment and other links, reduce players' own operation and maintenance costs, but this also means that the cost of use rises (for example, the processing of each 1GB of data may cost more than $8,000).
What if you choose an open source model? Software players can build their own large language models, which will make the capital investment more flexible and controllable, and can achieve control of data security, the shortcomings are of course obvious, the model iteration speed is slow, the efficiency is not high, after all, there is no free lunch in the world.
Who to choose in the end, the author's view is to combine their own reality, in the dimensions of funds, data, application scenarios and other dimensions to evaluate whether "people have me, people have me excellent", if there is no diamond diamond, directly choose the closed-source model may be a good choice.
Second, basic + application, it is difficult to do both
If we classify AI software, it can be divided into application software and basic software.
Because the know-how of the two tracks is almost completely different, the so-called more like a mountain, like the application field, we may be more familiar, for example, Microsoft has officially installed OpenAI's GPT-4 model into the Office suite, and the domestic "Microsoft" Kingsoft Office has also followed.
Today, we are mainly focusing on basic software.
Basic software is simply to provide a platform and tools for the operation of algorithms, and basic software can be divided into information security, data management, software development and other subdivisions.
Recently, the European Parliament just passed the Artificial Intelligence Bill, which is also expected to be the world's first bill on artificial intelligence. This means that in the future, large computing power models may be subject to stricter regulatory standards.
In addition, data as one of the future productivity factors, data screening, training and security will also be widely regulated and emphasized. In this way, data and information security players in the basic software industry will play a more important role in the AI track.
Major players in the AI monitoring software market
Third, learn from other people's children
It may be more abstract to say this, but the author will give a simple example.
For example, Palantir Technologies (software company, hereinafter referred to as P company) in the United States, mainly in big data analysis, the author after analysis, I feel that its core competition is: there is no general tool software, more than thirty percent of its products need to be customized, that is, to provide professional and customized services for the downstream.
Behind the customization, we dig a layer deeper, which means that we can improve data security and privacy protection capabilities.
From the performance point of view, in the first quarter of this year, its revenue has exceeded 500 million US dollars, a year-on-year increase, which is already valuable for a start-up.
Global AI market size
Although AI basic software provides a supporting role for the entire industrial application of AI, there is still a certain gap in the activity and actual progress of domestic players on this road, mainly concentrated in start-up software. Although many companies have an entry on "artificial intelligence basic software development" in their main business, not many actually disclose significant progress.
So in this blue ocean track, players may still have a long way to go.
Note: This article does not constitute any investment advice. The stock market is risky, and you need to be cautious when entering the market. There is no harm without buying and selling.