Since ChatGPT was released last November, the mini industry has broken the overall downward trend in the technology industry. Almost every week, “generative” AI (artificial intelligence) based on “big models” (large and complex algorithms that give intelligence to AI such as ChatGPT) is released.
On February 24th, Facebook’s parent company Meta released a model called LLaMA. According to reports, Tesla and Twitter boss Elon Musk also want to create an AI that is not as vigilant as ChatGPT.
There is also a model maintained by British entrepreneur Ben Tossell, which has just added Issaac Editor (“Isaac Editor”, which can help students write papers) and Ask Seneca (“Ask Seneca”, which can answer questions based on the writings of Stoic philosophers).
ChatGPT, which has 100 million users, is probably the most talked about and talked about by people. Mr. Tossell’s database reveals that the actual actions of generative AI exist in various forms of non chat services enabled by large models.
Each large model is trained through text, images, sound documents, or other data. This allows them to understand natural language instructions and give feedback in the form of text, art, or music. Although such systems have existed for some time, it is still necessary to rely on consumer oriented services such as ChatGPT to capture the imagination of the world – and investors alike.
As Mike Volpi of Index Ventures, a venture capital firm, pointed out, after being hurt by the cryptocurrency crash and the collapse of the original universe, he and his tech investors were just looking for the next high growth goal.
In addition, the big model makes it easy to create new services and applications on web browsers and smartphones (and even not limited to this).
Steve Loughlin of Accel, another venture capital firm, said, “You can open your laptop, register an account, and immediately start interacting with the model.”
Money is pouring in. According to reports, in addition to its early $1 billion investment, Microsoft invested $10 billion in OpenAI, a startup behind ChatGPT, in January this year.
According to Pete Flint points from NfX, another venture capital firm, there are now over 500 generative AI startups. Even without OpenAI, the total funding of these companies currently exceeds $11 billion. Mr. Volpi called it the “Cambrian Explosion.”.
So which generative AI platform can make a lot of money?
This is a question that has made the tech world scratching its head. Martin Casado of Andreessen Horowitz, a venture capital firm, and colleagues recently wrote on their blog that “it is not clear whether the industry will develop a long-term, winner take all situation.” Many startups offer ideas that emulate others, more like features than products. Even resource-intensive large models can eventually become low margin goods: OpenAI’s proprietary intellectual property rights, GPT-3.5, still lead, but other open source competitors are not far behind.
Generative AI is also falling into legal minefields. These models often get things wrong and may even overturn.
Microsoft is using OpenAI technology to develop a chat robot called Sydney for its Bing search engine. It has abused some users and has shown love to at least one person (since then it has been more tightly regulated).
AI platforms are also plagued by legal barriers to social media platforms. The copyright owners of online content used for existing model training are struggling against unauthorized or compensated content.
Photo gallery Getty Images and many individual artists have filed legal proceedings against AI art generators from companies such as Stable Diffusion. Stable Diffusion stated that it would “take these situations seriously, is reviewing the documents, and will respond accordingly.”
The news media are also afraid of AI grabbing its text content.
OpenAI has begun to make a low profile approach to the widely watched GPT-4 model upgrade release later this year.
This cannot suppress the desire of venture capital companies and others for generative AI. For risk averse investors, the safest investment target currently is suppliers with sufficient processing capabilities for training and operating large models. So far this year, Nvidia’s stock price, which is designed for AI application chips, has risen 60%.
Cloud computing service providers and data center landlords are also rubbing their hands.
No matter which AI platform can take the lead in the end, there is no mistake in selling chisels and shovels in the gold rush.