OpenAI Fundraising, LinkedIn AI Scraping, Claude Mini Performance
Key Points
- OpenAI is demanding $250 million minimum checks from venture‑capital investors for its next fundraising round, hinting at a potentially massive raise that could approach $100 billion.
- LinkedIn has added a hidden “generative AI data collection” toggle that defaults to on, allowing the platform to scrape users’ professional content for AI training without explicit consent.
- The lack of clear consent for large‑scale data harvesting raises serious privacy and accountability concerns, especially in the U.S., where users must actively opt‑out to protect their data.
- Anthropic’s Claude 1 mini can match the performance of the larger Claude 1 preview when given extra processing time, showing that latency can offset smaller model size in many use cases.
Full Transcript
# OpenAI Fundraising, LinkedIn AI Scraping, Claude Mini Performance **Source:** [https://www.youtube.com/watch?v=Vygt74Oyntg](https://www.youtube.com/watch?v=Vygt74Oyntg) **Duration:** 00:04:44 ## Summary - OpenAI is demanding $250 million minimum checks from venture‑capital investors for its next fundraising round, hinting at a potentially massive raise that could approach $100 billion. - LinkedIn has added a hidden “generative AI data collection” toggle that defaults to on, allowing the platform to scrape users’ professional content for AI training without explicit consent. - The lack of clear consent for large‑scale data harvesting raises serious privacy and accountability concerns, especially in the U.S., where users must actively opt‑out to protect their data. - Anthropic’s Claude 1 mini can match the performance of the larger Claude 1 preview when given extra processing time, showing that latency can offset smaller model size in many use cases. ## Sections - [00:00:00](https://www.youtube.com/watch?v=Vygt74Oyntg&t=0s) **OpenAI's Billion-Dollar Ask, LinkedIn AI Data Grab** - The speaker warns that OpenAI now demands VC checks of at least $250 million for funding, LinkedIn is silently enabling non‑consensual generative‑AI scraping of users’ professional data, and he also teases a discount for his AI Maven course. ## Full Transcript
three big pieces of news and a note on
how to get a discount on my AI Maven
course news number one open AI is able
to dictate any terms they want to get
people to invest in them and maybe you
and I aren't surprised but even I was a
little bit surprised by how high those
terms are for Venture capitalists they
are saying if you are not willing to
write $250 million checks you shouldn't
be talking to them their next
fundraising round will have at a minimum
a quarter of a billion dollar checks in
the round which shows how big that round
is going to be they haven't announced
the final size of it they're just
starting the Rays but I would
confidently expect it to be maybe close
to hundred billion dollar it's going to
be a big
raise number
two LinkedIn has decided to model
non-consensual generative AI scraping
they're not asking for your consent they
are sticking a new setting inside their
menu that says generative AI data
collection preference or something like
that I found it last night and they're
defaulting it to on without telling you
and so they're taking the largest
private professional data pool in the
world where like you and I are our
professional selves and we're writing
our content and they're saying all of
that is ours to collect and train our
model on but we don't have to tell you
that we don't have to ask for your
consent we can just do it now my
understanding is because of gdpr folks
in the EU don't have to worry about this
but in the the US I would go into your
settings and I would have a look and
make sure you are choosing intentionally
whether to share that data or not it's
not acceptable for generative AI
companies not to explicitly ask for
consent when they make a change that big
and I think it's important that we model
a degree of expectation of
accountability that that's not
acceptable it's not okay you need to ask
if you're going to collect people's data
for the purpose of training an llm it's
a norm that we need to establish
finally I want to call out one of the
interesting reasons why 01 mini is
performing as well as 01 preview it
turns out when you let 01 mini think for
a little bit it does just as well almost
as 01
preview in other words extra time when
you send in an utterance for it to
respond
to is just as good as extra size on the
model because 01 Min is a smaller model
and for most of us for most of llm
History our understanding has been
certainly my understanding has been that
if it's smaller it's generally not as
good now you can argue it's a smaller
model but it's specifically tuned and is
better at this use case absolutely bu
that but this is a general model and
it's still using the thinking time to
perform almost as well as 0 and preview
and you can see that in the chatbot
Arena uh scores you have two big scores
that are ahead of everybody else right
now they're like up here in the uh past
1300 1350s and then down here in the
1200s is every other model on the planet
including Chad GPT
40 and 01 mini is performing that well
because it's taking a little bit more
time to think and that's called log
linear Computing which is a mouthful to
say but basically amounts to the idea
that you can actually get those gains in
intelligence from a larger model by
giving the model time to respond I'll be
talking more about this in my lightning
lesson which is what I wanted to call
out it's one of the key differences with
the 01 models it's something we need to
think about as far as the kinds of
problems we give models and what we
expect from
them and uh you can sign up for it I'll
throw a link down below if you want to
Discount early also sign up for the wait
list the discount code is dropping on
Monday September 23rd and you want to be
there it's a decent discount code for my
mavin course so to recap lightning
lesson will cover 01 it's free 30
minutes sign up and the wait list which
is a nice discount right now on my Maven
course is dropping on September 23rd and
the reason why you might want to jump on
that is because I've had a ton of
interest I'm already up to almost 50
signups for the course uh I I want to
make sure I have like appropriate
instruction for everyone and so I
actually don't know how big I can get
the course before I have to cut it off
so if you're interested in the course
you should hop in so that I can be sure
that I can
provide you the kinds of instruction on
AI that that you want that I love to
give and that will help you accelerate
your career because that's ultimate
ultimately what this is all about is how
can you accelerate your career