Partner
@RadicalVCFund
, AI columnist
@Forbes
. "the machine does not isolate man from the great problems of nature but plunges him more deeply into them."
Venture capital requires you to be exceptionally optimistic and exceptionally skeptical at the same time. It is not an easy balance, but man it is fun.
For those of you intrigued or confused by
@ylecun
and
@GaryMarcus
’ ongoing Twitter spat about symbolism and connectionism, check out my recent Forbes article for some context:
It was an absolute pleasure speaking at Upfront Summit last week with
@DaveRogenmoser
from Jasper,
@c_valenzuelab
from Runway, and Phil Blunsom from Cohere about generative AI. One sharp group!
Nice shoutout in the New York Times today.
Tech hype cycles come and go. We
@radicalvcfund
were committed to AI before this one came, and we will continue to be after it passes.
Great line from
@GaryMarcus
’ new book: “The real reason computers can’t read is that they lack even a basic understanding of how the world works.”
This is why the Turing Test (while imperfect) has stood....
A lot of people are talking about conversational search.
just shipped it.
Huge congratulations to
@RichardSocher
and
@YouSearchEngine
— the team creating the future of search!
Today, youChat goes live.
Open, broadly capable, conversational AI for search with knowledge of recent events and citations of sources.
Search and chat of the future:
Thrilled to welcome Dominic Barton (former head of McKinsey) and Aaron Rosenberg (former head of strategy & operations at DeepMind) as Partners at Radical!
LLMs have ushered in a period of unreal possibilities. There are dozens of great startups that should exist that don’t yet. We have a ludicrous oversupply of great startup ideas.
Great ideas are 1%. Execution is 99%.
The next couple years will be fun!
It was a real pleasure speaking at the SALT Conference in NYC last week about all things AI alongside
@AndrewYNg
,
@May_Habib
and
@1vnzh
. Very fun conversation!
@aifazza
@GaryMarcus
@lacker
if GPT-4 has 10x or 100x more parameters than GPT-3, I agree that error rate will be lower. expect diminishing returns, though.
more broadly, this doesn't change the fact that it still will lack internal representations, understanding, the ability to reason 😊
I have published AI predictions in Forbes for the past four years.
These are my spiciest predictions yet 😏
Let me know what you agree with and what you disagree with!
There is a lot of buzz right now about LLMs and foundation models.
@twelve_labs
is building an entirely new kind of foundation model: a foundation model for videos.
Excited to lead Twelve Labs' latest round with our friends
@IndexVentures
and
@WndrCoLLC
Did you see the news? Today Domino announced a new ML model monitoring product, $43M in new funding, and innovations to our enterprise data science management platform.
#modelops
#modelmonitoring
#datasciencemanagement
Congratulations to Radical portfolio companies
@CohereAI
, Waabi and
@UnlearnAI
on being named to Forbes' "AI 50" list of the 50 top AI startups in the world.
It was a lot of fun serving as a VC judge for the list this year!
The AI 50 2023 ()
It is a real pleasure to be giving a TED Talk next week at
@TEDAI2023
alongside Andrew Ng, Reid Hoffman, Ilya Sutskever and so many other amazing AI leaders. Looking forward to it!
@ylecun
this is a straw man. I do not know of anyone calling for a pure rules-based approach. Advocates of a hybrid approach, like
@GaryMarcus
, are suggesting combining learning-based and symbolic techniques. Learning remains essential — this is inherent in the word “hybrid”
This is one of the really big, outside-the-box ideas in AI today: should machine learning models have more intelligence "built in" (like humans do)?
@TonyZador
PSA: Building cutting-edge large language models is very hard.
There are very few people in the world who know how to do it.
If a magical shortcut appears that suddenly makes something very hard seem very easy, it is probably too good to be true.
to those into genetics: it is truly incredible that Darwin and Mendel lived at the same time on the same continent and never crossed paths. the ultimate “what might have been”
It's not often that two of the world's leading legends get on a stage together to talk about the past and future of their field.
This conversation between
@geoffreyhinton
and
@drfeifei
, brought to you by
@radicalvcfund
, will be a special evening.
It's been an insane year in AI. To try and understand what's going on, I asked 10 awesome investors what trends and companies to watch out for.
Here's what they said 👇 🧠
🔗
awesome AI angel portfolio from
@olivercameron
- exciting set of companies that showcases the transformative impact that AI will have across industries
AI is changing everything, and it's not just ChatGPT or self-driving cars.
You'd be surprised at just how broadly AI is being applied across tons of industries.
I’ve now invested in 50 AI startups, and here are some recent investments using AI in fascinating applications 👇
It was a lot of fun speaking with
@paulinebhyang
at Altimeter and
@danny_lange
at Unity about all things AI on the inaugural episode of the Cross Validated podcast. Thanks for having me!
Excited to launch Cross Validated podcast, where we speak with practitioners & builders making AI deployments a reality!
@altcap
Our first guest is
@danny_lange
, SVP of AI
@unity
, w guest host
@_RobToews
@radicalvcfund
We have been studying and investing in AI over last few yrs
if you work in the field of artificial intelligence, and you don't find yourself bumping up against philosophical questions pretty regularly, you're doing it wrong
It was a blast guest-hosting Redpoint’s AI podcast alongside my pal
@jacobeffron
and interviewing
@npew
, who leads OpenAI’s commercialization and product efforts.
Fascinating and enlightening discussion. Enjoy!
Excited about our multi-year partnership to power
@CohereAI
’s natural language processing technology with Cloud TPUs, making it easier and more cost-effective for any organization to use AI to better understand their customers.
“Any intelligent fool can make things bigger and more complex. It takes a touch of genius and a lot of courage to move in the opposite direction.”
—Albert Einstein
Hubert Dreyfus’ classic book “What Computers Still Can’t Do” (originally published in 1972, updated in 1992) is a great reminder that, while the amount of compute and data available today is a new thing, the fundamental concepts at the core of AI are not
Some choice quotes 👇
in a world that has gone crazy over AI, I am deeply appreciative to work with the best people in the world at the biggest VC firm in the world focused exclusively on AI. onward!
It is sometimes said, “Today’s AI is just statistics.” The only part of this that is incorrect is the word “just.”
Statistics is incredibly powerful, especially when paired with digital computation — more than powerful enough to generate intelligent behavior.
Good article on LLMs at Forbes.
The media are starting to agree with my much-criticized statements about LLMs.
"LLMs as they exist today will never replace Google Search. Why not? In short, because today’s LLMs make stuff up."
great reporting by
@theinformation
and big ups to
@DominoDataLab
! the "AI boom" has just begun, and Domino is in a great position to turbocharge it in the enterprise in the years ahead
it's not often that you see an AI startup develop an entirely new AI model architecture, purpose-built for its needs. huge congratulations to
@charleskfisher
and the
@UnlearnAI
team - one of the most original and imaginative AI companies out there, in any sector!
Are you tired of same old generative model architectures?
Do you long for something new and intriguing?
Then we've got you covered ...
Introducing Neural Boltzmann Machines!
With the powers of DNNs and RBMs combined!
🧵 (w/ paper and code)
The full
#Horizons
issue on
#AI
is worth reading:
1. The AI Opportunity - written by GPT-4😉
2. The AI Threat - by Nouriel Roubini
@Nouriel
3. From Shocking to Mundane - John D’Agostino & Manolis Kellis
@JohnJDAgostino
@ManolisKellis
4. The Aliens Have Landed, and We Created Them
Huge congratulations to the team at
@DominoDataLab
! Domino is all substance and no hype, building one of the most meaningful companies in the burgeoning field of MLOps. Onward!
excellent, thoughtful piece from
@molwelch
about what is needed to productize and operationalize generative AI at scale. if you're excited about generative AI, read it!
lots out there on application layer opportunities in generative AI (and, of course, how crazy good chatGPT is!), but less on what’s unsolved in this new paradigm.
new piece on what’s missing in generative AI, where I cover: 👇
I’m as big of an AI bull as there is and have been for about ten years but holy shit most of the biggest grifters I know have become “AI people” and it absolutely terrifies me.