1. When will
the new AI/LLM features be available?
In Summer 2025, probably in the second half of July. The LLM integration is
already working, as you can see from the sample queries (PDF,
video). But there are two main
things that still need to be done. First, we need to add more functionality to save and
retrieve your AI-based queries. Second (and more importantly), we need to set up the
system that will allow you to purchase "AI credits" for these queries (see #2
below). Another reason is that I will have three international trips in the next 2.5
months (including keynote address in
Spain and
Germany), which accounts for the majority of the delay in releasing these
features.
But it has taken me only about 1.5 months to add all of the AI/LLM functionality
that is currently available, so July 2025 should be a reasonable goal.
2. How will the "AI credits" work?
First, it is important to
realize that every time we send an "API call" to an LLM (GPT,
Gemini, Claude,
etc), the LLM charges us for that access. Imagine that each of the
75,000
monthly users do just 12-15 LLM-based searches each month (and some
people will do many more than that). That would be more than one million
API calls to the LLMs, and this would cost us thousands (or tens of
thousands) of dollars each month. As you can imagine, it's not realistic
for us to pay that cost by ourselves; the corpora would cease to function
within a few months.
So the plan is to have people
purchase "AI credits" in advance -- for as little as $4.95. Each time they do an
AI/LLM-based query, it will deduct $0.01 or $0.02 from their credits (depending
on the type of search and how much data is sent to and received from the LLM). So they will have
anywhere from 250 to 500 API-based queries (if they have paid $4.95) before they
need to purchase more credits. And the corpus interface will let them know at
all times how many credits they have left. Payment will be made by credit card at PayPal, as
is done currently for the premium licenses
(with the option of payment by AliPay or WeChatPay for users from China).
Note that these AI credits will be
separate from (individual) premium licenses. In other words, if you already have
a premium license, you would still need to purchase the "AI credits". But you
can also buy the AI credits even if you don't have a premium license -- they
are two completely separate things.
And most importantly, the AI
features (via API access to the LLMs) are completely optional. If you
want to keep using the corpora as you have in the past (without the insights
from AI/LLMs), that would be perfectly fine.
3. What about
users of academic licenses?
Academic licenses are
"university-wide"; they include all of the users at a university. But again, suppose
that there are 500 users from a university, and each person does an average of
just 40-50 AI-based queries each month. After a year, that would be 300,000
queries (500 x 50 x 12). If English-Corpora.org had to pay for all of that API
access itself, it would actually lose money for the academic licenses for
some universities.
But it also doesn't make much sense for a university to pay a
"surcharge" ahead of time, which could be used by anyone from the university,
since a handful of people might use up all of those AI credits themselves, and
leave nothing for others. It also doesn't make sense to have a university pay
for the API queries "after the fact" -- after users at their universities have
incurred charges for their API access.
So even though it's probably not an optimal solution, we will probably have individuals who are
part of a (university-wide) academic license pay for their own "AI credits". We
know that this is a change from the past, where everything was paid for by the
university. But it does seem like the only workable solution. (If you have other
ideas, please let us know: admin@english-corpora.org; we're still trying to work
this out).
