I think this is at least the third comment I have seen recently complaining about "AI" APIs. As a non-developer, this is difficult for me to understand.
I have had what appears to be the same, or similar, complaint against "Web APIs" for many years when trying to access public iinformation. Not that long ago, websites did not have "APIs". Generally, I still extract information from the public websites rather than use "APIs". This requires no sign up and is guaranteed to work as long as the website exists.
Before "Web APIs", websites did not routinely collect email addresses, track usage, rate limit individuals and charge subscription fees in exchange for access to public information.
Sometimes the "Web APIs" are free but "sign up" is always required. There is email address collection, usage is tracked, "accounts" are rate limited. In the past, some HN commenters also complained that these "APIs" are unreliable in the sense that they can be "discontinued" suddenly for any reason. Anyone depending on them has no recourse.
These "Web APIs" became so common that their rationale went unquestioned. For example, why not let www users download bulk data. In rare cases, this is an option, e.g., some government websites, Wikipedia dumps, etc. But this is the exception not the norm.
In light of these comments complaining about accessing LLMs through Web APIs I am wondering:
Are "Web APIs", "SaaS", etc. now being turned against the people who concertedly promoted these tactics, namely, software developers.
I always saw "APIs" as an easy way website operators could deny access to www users. Like some requirement to have an "account" in order to access public information. Those providing these "APIs" have the data in a format they could provide for download, e.g., JSON, but bulk downloads are not provided. This tactic of charging fees is quite different from how websites operated for the first decades I used the www.
Similarly, LLM providers have details about the models, e.g., weights, etc., in formats they could provide for download. Instead, users are encouraged to "sign up", "subscribe", or whatever, for access to public information^1 through an "API".
1. Assuming the provider obtained the training material from the www, a public resource.
I have had what appears to be the same, or similar, complaint against "Web APIs" for many years when trying to access public iinformation. Not that long ago, websites did not have "APIs". Generally, I still extract information from the public websites rather than use "APIs". This requires no sign up and is guaranteed to work as long as the website exists.
Before "Web APIs", websites did not routinely collect email addresses, track usage, rate limit individuals and charge subscription fees in exchange for access to public information.
Sometimes the "Web APIs" are free but "sign up" is always required. There is email address collection, usage is tracked, "accounts" are rate limited. In the past, some HN commenters also complained that these "APIs" are unreliable in the sense that they can be "discontinued" suddenly for any reason. Anyone depending on them has no recourse.
These "Web APIs" became so common that their rationale went unquestioned. For example, why not let www users download bulk data. In rare cases, this is an option, e.g., some government websites, Wikipedia dumps, etc. But this is the exception not the norm.
In light of these comments complaining about accessing LLMs through Web APIs I am wondering:
Are "Web APIs", "SaaS", etc. now being turned against the people who concertedly promoted these tactics, namely, software developers.
I always saw "APIs" as an easy way website operators could deny access to www users. Like some requirement to have an "account" in order to access public information. Those providing these "APIs" have the data in a format they could provide for download, e.g., JSON, but bulk downloads are not provided. This tactic of charging fees is quite different from how websites operated for the first decades I used the www.
Similarly, LLM providers have details about the models, e.g., weights, etc., in formats they could provide for download. Instead, users are encouraged to "sign up", "subscribe", or whatever, for access to public information^1 through an "API".
1. Assuming the provider obtained the training material from the www, a public resource.