Google is getting ready to release its own competitor to Chat GPT. There are many possible ramifications. Small startups now will face firms far better capitalized, with more scale and developed ecosystems of developers, users, customers, roles and business partners.
As always is the case, that means many smaller firms will be acquired. Some will simply disappear. Since all generative language engines analyze lots of text, Google should have advantages. It indexes an awful lot of text. The adage that artificial intelligence models benefit from huge datasets is apt.
Other questions are harder to answer and assess. In principle, generative language engines should find common use as an alternative to use of search engines. So it is not surprising that Google and other search engines will try and marry generative text capabilities to their existing search platforms.
Just how far the capability might spread in various industry verticals likewise is hard to assess, sometimes because there are legal issues to using the generated text. In other cases, such as health care, the generated text can help point users to other text that can be used in care, without creating ethical or other liabilities.
We might find that is the pattern for many other industries: generated text might be useful for high-level backgrounding but not useful as a tactical guide to problem identification or solutions.
On the other hand, in contained use cases, such as customer service, AI-generated text might be quite useful as a substitute for human action. Answering generic questions for which there are structured answers seems easy. Generative text becomes less reliable as the range of possible answers, in context, is required.
If you have used engines such as ChatGPT, you already know the current state of the art is that high-level summaries can be quite useful. But detailed, in-context tactical information is another matter.
The other observation is that artificial intelligence now is more rapidly entering mainstream use after many decades of gestation. Useful precisely for allowing conversational answers to queries, generative text engines should quickly be useful for queries that have structured answers.
But many questions that must include many value assumptions will always have multiple answers, depending on context and point of view. In such cases we might find generative text of some value, but without the context on training models or degree of dispute about “facts", we might not be able to “trust” the generated text.
No comments:
Post a Comment