AI Breakthroughs: Cancer Detection, Legal Fight, Funding
Key Points
- A new AI‑driven cancer‑detection platform called “CHEF,” built on a transformer architecture, claims 96 % accuracy across 19 cancer types and can even flag novel survivability traits from uploaded pathology slides.
- A Massachusetts family is suing their school after the child received a D for using AI on a social‑studies assignment, sparking a legal debate over whether AI‑generated work constitutes plagiarism or the student’s own intellectual property.
- Perplexity AI, the AI‑powered search engine known for rich text summarisation, has raised $500 million at a $9 billion valuation, expanded into internal and external file search, and launched new financial‑charting tools that some fear could challenge traditional services like Bloomberg.
Full Transcript
# AI Breakthroughs: Cancer Detection, Legal Fight, Funding **Source:** [https://www.youtube.com/watch?v=xHloxRZV3gA](https://www.youtube.com/watch?v=xHloxRZV3gA) **Duration:** 00:06:26 ## Summary - A new AI‑driven cancer‑detection platform called “CHEF,” built on a transformer architecture, claims 96 % accuracy across 19 cancer types and can even flag novel survivability traits from uploaded pathology slides. - A Massachusetts family is suing their school after the child received a D for using AI on a social‑studies assignment, sparking a legal debate over whether AI‑generated work constitutes plagiarism or the student’s own intellectual property. - Perplexity AI, the AI‑powered search engine known for rich text summarisation, has raised $500 million at a $9 billion valuation, expanded into internal and external file search, and launched new financial‑charting tools that some fear could challenge traditional services like Bloomberg. ## Sections - [00:00:00](https://www.youtube.com/watch?v=xHloxRZV3gA&t=0s) **AI Cancer Tool & School Lawsuit** - The segment highlights a new transformer‑based AI system called Chief that detects 19 cancers with 96% accuracy, and a Massachusetts family suing a school after their child received a failing grade for using AI on a social‑studies assignment. ## Full Transcript
four pieces of AI news that you might
have missed we're going to start with
good news today there is a new AI cancer
detection tool it works across 19
different types of cancers goes by the
name chief for reasons I don't fully
understand and it is apparently 96%
accurate at detecting cancers across
these 19 different types it also detects
novel survivability characteristics
which I'm not exactly sure what that is
but it does sound positive so I think
one of the interesting things is that we
are starting to see this concept of a
transformer-based architecture work
across many many different applications
in this case it's a chat GPT like
interface that allows you to upload a
slide of a cancer tumor and then the
Transformer architecture will look
through the pattern recognition based on
the large training data set of cancers
and come back with a response and that
core Transformer based attention-based
architecture
is very very good at cancer detection
just like it's good at language just
like it's good at image generation it's
a it's it's this is a remarkable time
that we're living through to have one
particular architecture in computer
science work across so many fields okay
that aside number
two there is a family in Massachusetts
that is suing their school because the
school gave the kid a d on an assignment
in social studies which is not by itself
a sueable offense but apparently the
child used Ai and there was no rule in
the handbook that said the child
couldn't use AI but on the basis of
using
AI the school is saying you get a d and
what's interesting is the school is
claiming that it's self-evidently
cheating to use AI it's
plagiarism and the parents are claiming
it's not that the AI generated report is
the child's property and the child's
work we will see what happens but this
is going to be one of those interesting
cases to watch because in school
settings plagiarism is obviously a
problem in most places plagiarism is
obviously a problem and at the end of
the day you either believe that the AI
is generative and is producing work in
conversation with you based on a large
training data set or you believe it's
tent to amount to cheating and the
courts are going to weigh in here and we
will see what they say so that's number
two number three the AI search engine
perplexity has decided that it is worth
three times what it was worth earlier
this year and is Raising $500 million at
A9 billion valuation up from three
earlier these are the guys that just
launched internal file search so
previously they were just doing doing
search engine like work like Google did
but with uh much more in in-depth
enriched High text
summaries now they are doing internal
file search they're doing the external
file search and they are coming for
financial charge they have some like
Financial stuff that they launched
recently where you can build Financial
charts and I saw people get very excited
and say that Bloomberg was in danger and
if you've ever seen Bloomberg no they're
not um one of the things that's
interesting is with perplexity they are
using very high intent keywords and so
you have 10 or 11 words in a typical
perplexity search and just two or three
in a particular Google search and so the
idea is they will be able to monetize
more effectively than Google will
because there's so much more information
in that initial prompt we will see but
that's the pitch that they're making to
the investors all right number four last
but not least Microsoft is launching AI
agents we said it was coming by the end
of the year here Microsoft is jumping
into the right they are launching it
with uh co-pilot they're allowing
businesses to build it first and the
application they gave I am not kidding
you it's a McKenzie
application and the application they
gave is that the agent can read the
McKenzie partner's email and figure out
who to forward it
to and that's their AI agent application
and they say this will save McKenzie
Partners 30% of their
time I don't even have words Mackenzie
partners are spending that much time on
email
McKenzie can't come up with this as a
way sort of to to work with a a tech
firm to to write this anyway because
you've been able to do this you just
have had to work on it yourself and roll
your own code like yes Microsoft now has
tools that allow you to do this with no
code that's great but you've been able
to parse
emails run them through an llm generate
responses and understand the semantic
meaning for a while it's just putting
the piece together that's taken a you
know a tech team so apparently McKenzie
hasn't had a tech team to do that hasn't
thought of doing it and has decided to
do it with
Microsoft and Microsoft is very happy
and has sort of given them a nice little
graphical user interface with a little
tree based architecture where you can
like almost like zapier right like just
Define what you want and it will
magically
work I think agents are going to feel a
little
bit they're going to feel a little bit
blah if this is the way they're being
introduced I think they have a ton of
potential I think we're going to see
agents which are much more exciting
soon but this is not the way to
introduce it agents that that just sort
of do the partners email at McKenzie
okay okay and I think they're going to
be real productivity enhancers so like
the boring agents are okay it's not bad
but from a marketing perspective you'd
want something with a little bit more
Pizzazz so we will see we'll see what
the Microsoft marketing team does they
probably won't listen to me and those
are your four pieces of news cancer
which is
good perplexity raising a triple the
valuation so they must be getting
increased usage we have the parents
suing for AI and we have Microsoft
launching AI agents but kind of picking
a boring use case let me know what
interesting use cases you've seen for AI
agents