Dario Amodei - CEO of Anthropic | Podcast | In Good Company | Norges Bank Investment Management
Вставка
- Опубліковано 25 лис 2024
- Dario Amodei CEO of Anthropic: Claude, New models, AI safety and Economic impact
How much bigger and more powerful will the next AI models be? Anthropic’s CEO, Dario Amodei, joins Nicolai in Oslo to discuss the latest advancements in AI, the economic impact of this technology, and the importance of responsible scaling. Dario also shares his excitement about the upcoming models and his thoughts on who will profit the most from AI in the future. Don’t miss this enlightening conversation about the cutting-edge developments in AI. Tune in!
In Good Company is hosted by Nicolai Tangen, CEO of Norges Bank Investment Management. New episode out every Wednesday.
The podcast In Good Company is part of Norges Bank Investment Management , a podcast about our investments. We are transparent about how the fund is invested. In this podcast, you to get a deeper insight into the companies the fund is invested in.
Our CEO, Nicolai Tangen, has in-depth interviews with the leaders of some of the largest companies in the world. You will get insight into their leadership principles, the company’s strategy and how they are dealing with a large investor like us. You will also learn more about our role as an owner of the companies.
You can read more about the podcast on www.nbim.no/en...
#AI #Anthropic #ingoodcompany #oljefondet #DarioAmodei #claude
I don’t say this about a lot of people but I really admire Mr. Dario Amodei. He appears to be a smart, honest, and a kind person. He changed my original perception about him which was a Chat GPT copy cat. I have used Claude for quite sometime now and would say that it is a superior model compared to everything else that is out there. What’s weird is that I feel like Claude is more friendly and tries harder than ChatGPT to understand me. I can’t wait for what Claude is going to be like in the future.
I hope he doesnt go the same way as sam altman and goes hellbent on scuffed releases and safety... although claude is already the worst in safety rn...
@@incriptionas Jan Leike said, with Sam Altman is all about shipping new shiny products at the detrament of all else.
I dunno... Amodei seemed cagey to me.
@atlasatendido rejects prompts way too much
@atlasatendidoit can’t even generate me a reply to send to someone. It just stops and rejects that it wants to push more authenticity 😂 Claude is good for coding but still falls short on so many areas
I think Nicolai operates on the ethos of "if you can't explain it simply you don't understand it" because he is great at getting people to break information down into its purest form .. great pod 👍
I would challenge him to explain why that is true in a way that a 2-year-old would understand. (Actually understand, not just parrot.)
While most day-to-day concepts CAN be described simply enought that a slightly-above-average can grasp some of it, there are plenty of topics out there that are really challenging to properly explain to an average person without any knowledge of the field, because they're built by stacking a dozen or more layers of knowledge on top of each other and/or because they require manipulating 7+ abstract concepts within a single thought.
The claim "if you can't explain it simply you don't understand it" really only works well for topics that are eiter relatively simple or within a relatively homogenous group, in terms of capacity and knowledge.
One could also make the claim: ' If you believe that "if you can't explain it simply you don't understand it" is a true statement, you've never understood something that is irreducibly complex'
@@haakoflo I see your point but respectfully disagree that the principle "if you can't explain it simply, you don't understand it" only works well for relatively simple topics or within a homogenous group.
High-level teachers, researchers, and CEOs often explain incredibly complex issues in simple, accessible manners. They can do this because they have spent extensive time studying these topics and refining their understanding. Nicolai’s goal isn't to oversimplify to the point of losing essential details but rather to make the core concepts comprehensible to a broader audience of his podcast (us). His manner of being, I believe, is fundamental to effective communication and education, as it empowers more people to engage with and understand the material, regardless of their background.
The concept of "layman's terms" exists for a reason, enabling complex information to be communicated without requiring specialised knowledge. I believe that Nicolai, from an investor's perspective, excels at breaking down the information his guests present into these more accessible terms. For example, while I may not fully understand the intricacies of quantum computing, I can grasp the basic concept that it's a type of computing that uses quantum bits to perform calculations much faster than traditional computers.
Nicolai must have heard thousands of pitches and asks in his time, and I believe this history and his high-level position have given him the confidence and ability to break down people's information quickly into "layman's terms." Simplifying complex topics doesn't mean stripping them of their essence but making them more approachable and understandable. This skill is invaluable, especially in fields where experts need to communicate their ideas to non-specialists.
@@pinkbluered7481 It's not that I think that there isn't a core of truth to the statement Nikolai is making. Most (if not all) concepts can be simplified a lot as one gains deeper understanding of it.
Someone who possesses a much deeper understanding of a topic than other experts of a field, and who also have excellent communication skills and the humility to avoid overly technical terminology CAN certainly make such topics available to a wider audience than those who would be able to follow the typical presentation of the topic.
I'm with him this far. However, there are two issues:
1) Some topics are VERY complex. Take for instance Quantum Chromodynamics. A brilliant communicator MAY be able to simplify that enough that bright undergraduate students of Math or Physics can follow most of the ideas. But I don't think it's possible to explain it to a median person, someone who struggled with Pre-Calculus, without gross oversimplification. There are are simply too many things underpinning the Quantum Field Theory of QCD that violate intuition, but that are needed to even start understanding QCD.
A brillian Physics communicator would have a hard enough time simply giving some idea of what undergraduate level Quantum Mechanics is about to such an audience.
2) Understanding comes at many levels. While deeper understanding DOES generally make it possible to simplify a topic, it doesn't meant that slightly shallower understanding is not also to "understand".
Furthermore, people have different ways of understanding. The mental concepts used to understand something are different from person to person.
I would argue that a person may understand something pretty well (more than well enough for their needs), without having the ability to present it in much simpler ways than are usually found in textbooks.
-----
Telling someone at this stage of understanding that they dont understand the topic is an unnecessary insult.
Still, there ARE many semi-complex topics out there that CAN be explained in simple enough terms that Nikolai indeed can understand them.
And for someone in his position, it's reasonable to demand that people do their best to communicate those topics in a manner that can reach him and the people on his team, at least if they want him to listen to them.
But if he thinks that he (or even someone much less educated/intelligent than him) could understand ANY topic, if it's only communicated by someone with a deep enough understanding, then it makes me think he's either not had much exposure to truely complex material or he hasn't had much exposure to average intelligence people (or both).
@@pinkbluered7481 you can explain anything in layman's terms, but you may be oversimplifying to the point where people's understanding is very shallow. I see this all the time with astrophysics and particle physics videos; in the comments people aka kooks propose their own stories like dark matter is aliens hiding in Dyson spheres or the universe curves back around in the time dimension, not realizing that the accepted theory is based on a vast amount of hard mathematics that has to fit fairly well with gigabytes of experimental data.
The same happens in AI explanations. People grab onto simplifications like statistical sampling or the most likely next output to confidently the AI is "just" doing pattern matching or autocorrect. But linear algebra in a million dimensions is unavoidably very different from all our commonplace notions. It's not like other things.
He is not rude.He never allows interviwee to digress from topic.He is my role model .I love his podcast
I trust Dario much more than Sam Altman with the future of Ai.
Sweet. My favorite podcast with a super timely interview. We still use OpenAI's models for mission-critical and secure enterprise scenarios; however, I'm enjoying Sonnet for ad hoc work.
I agree that the Anthropic models feel "warmer" which makes them better for writing.
I was also doing some SQL queries this morning for analytics, and Sonnet inferred my intention and gave me a little more data to make my results more useful compared to GPT-4o and GPT4. You could argue that it took the liberty of extending my intentions instead of giving me exactly what I asked it to do. However, the final results were more useful. Later in the day, Sonnet bombed on another log query task, GPT4 (classic) also failed, and GPT4o got it right.
Thanks to all the companies pushing forward. It's a real pleasure to live through this transition.
Well-stated. Claude seems to have a different approach that seems almost more “empathetic.” Even if it’s not always better, it does seem to take a different approach than the competitors.
@@anothername2730 Dario seems to be quite friendly, just like Claude. Dario is actually kinda cute, too.
Terrific interview with an important figure in the industry.
Holy shit, had no idea Dario worked on Concrete Problems in AI Safety. Absolute legend.
The model he refers to that is best in world is 3.5 sonnet. It’s true. It is amazingly sharp. Better currently than chatGPT
Kudos to the Anthropic team for trying to understand how the model works in order to prevent any future harm. Very cool forward thinking.
This episode is so timely and interesting. In Good Company is one of the podcasts that I never miss any new episode from. Thank you for making these available for the world!
Liked both the interviewer and the interviewee. Good questions and good insights without drama 😂
Pioneer of advanced ai: Dario Amodei
And He has just started but i can already see the headlines
Thank you Dario for disrupting your vacation for this
Great interview. I've really come to respect Dario. He seems to have a really clear and positive vision while still being grounded in both risks and business realities.
Watching this video two months later and o1 just launched. Almost as he inferred, o1 gives you the step by step thinking process insight as well. He was so right about it.
Dario is a pure genius. Can't wait to see what Claude 3.5 Opus does by end of summer
Great interview. Great insight. I thought the interviewer was really good and had very thoughtful questions even beyond the company. I can tell Dario speaks from the heart. 😊
56:00 Bingo! This resonates so strongly with me. So well done, please continue on this journey of connection. You give me hope for our future with this quality of leadership. 🙏👍
Thank you for this interview. Loved the questions and conversation.
This channel is worthy of a million subscribers. High level interview with high level highly interesting people of our time.
Great interview. The analogue that todays models are like undergrads is so interesting and terrifying at the same time. Personally I have gone from making black jack console games with GPT3.5 to doing frontend aps with GPT4, and now with Claude 3.5 Sonnet you can do 2D simulations of the solar system with accurate trajectories and all with no bugs after just one prompt. I’m sure this progress will continue during the next months and years.
08:25 Dario Amodei discusses the latest breakthrough in AI models and the importance of interpretability in understanding AI decisions.
16:50 Entropic's long term goal is to drive the industry towards ethical and responsible AI practices.
25:15 AI models becoming more efficient and powerful.
33:40 Regulating AI and its impacts worldwide
42:05 AI advancements can lead to groundbreaking discoveries in biology and drug discovery.
50:30 AI's impact on narrowing rich-poor gap and avoiding closed loop ecosystems
58:55 Importance of democratic coalition in AI development
1:07:15 Decentralized innovation vs. collaborative effort
Crafted by Merlin AI.
It is truly amazing to see the work you are doing, providing access to these incredible people and their valuable perspectives on today's most important issues. Thank you so much.
Great guest. Great questions. Thanks!
great! dario's face/eyes keep moving, rather direct into eyes. love it.
Great interview!
I like the discussion of looping in non-tech people as users. Wouldn't it be cool to see a collab between Anthropic and Khan Academy or Brilliant?
Claude is apperantly higher IQ than other models. its so sattisficating to work with sonnet 3.5 it beats everything that I tried.
Can’t wait to run Claude locally using MLX
Nice. Really nice.
So informative
at 14:40 his comment about nvidia is super important. He stressed on the fact they are not the only player anymore.
admin you have not uploaded the second Jensen huang podcast to youtube?
great guy
Have been watching a ton of these pods and have been really liking them. Nicolai was pretty rude in this one. ‘Did you have any friends growing up?’ Jeez man
❤
I feel safe with Dario.
I feel the same about all of them really. Not even Zuck concerns me, and I have been very against Face as a product. I don’t know what’s going on in china, but the western players all bring me great confidence.
Sam Altman? I don't trust him enough.
@@Herr2Cents Same. Especially after Ilya's departure. Sam's ambitious approach, while great for startups, might not be the best fit for revolutionary technology with significant implications. We can't afford to 'move fast and break things' with AI.
Weird that it has only 13k views, but in fact Dario did not reveal anything extraordinary
Dario seems to be quite friendly, just like Claude. Dario is actually kinda cute, too.
I laughed about the apologies from the AI bots. Claude does it as well almost in any response it makes a mistake.
But it seems better than open AI
Interesting that Ilya couldn't align himself with Anthropic.
I trust Dario Amodei more than Sam Altman of OpenAI.
The last name itself tells all:
“Amo_dei” in Latin or Italian it means: “love_of_God”
"like" counter should be at certain parts. @ 37:13
timestamps please. we are all accelerating and might not have time for an hour long interview
Interviewer asks so many questions about things no one is going to be able to answer.
Please make timecodes.
They just keep talking in the same circle. Regulation needs to happen, but no one wants to put the regulation in place because it might stifle progression, but the progression is what's causing the alarm for additional regulation.
So, I guess what we can expect is that we will just have to keep rapidly adapting to the releases from these companies.
I would just prefer it if they said it outright as opposed to attempting to soften the blow.
Why is the interviewer being rude?
he is scandinavian lol, not rude. this is normal there, how people talk (very directly)
@@Sindigo-ic6xq Nah just rude
As a Scandinavian, I can confirm that he is not being rude by Norwegian standards; just direct.
I'm now avoiding new purchases to avoid slipping into a bear trap. However, I'd like to know where the best investment opportunities are during a downturn. My goal is to retire comfortably on around $1.2 million.
Wishing you luck! The key to a happy retirement is to live the life you choose, if you have a certain amount in mind for your retirement, it is only appropriate to work with a knowledgeable advisor to make plans.
I concur that guidance should not be disregarded but rather followed. I remember a few summers ago, not long after my horrible divorce, I was in dire need of advise on investments to keep my head above water. Fortunately, I met someone with grit who greatly assisted me in growing my reserve, which despite inflation has increased from $360k to about seven figures now.
nice gains there! Please share your advisor's info with me. Due to my increased income from work, I'm thinking about investing in stocks and other assets that will enable me to gradually build wealth
Leah Foster Alderman is the licensed expert I use. Just look up the name.
You would discover the necessary information to deal with a correspondence to schedule an appointment.
Drop the model THEN retroactively farm hype, OpenAI should take notes
Race to the top doesn't happen if language is incompatible for countries.
What happened to the previous version of this video?
What was in the previous version ?
Is this the new conspiracy?
"The Dario interview is back up and they cut these 33 seconds where Dario clumsily "no comments" about whether Claude will be on Apple devices."
@@TheLegendaryHacker thank you 💯
33:48
Since when is an exponential smooth?
Society cannot deal with exponentials. Neither with pandamic corona-like exponentials, nor with AI exponentials.
its so sad to know that i cant invest this company.
Mr. Amodei- please have your chief of staff contact me at your earliest convenience. Thank you.
imo Nicolai was pretty rude to Dario by cutting him off repeatedly and ignoring him at points. I know you want simple answers but let the dude speak.
if found the interview really good with many questions that have not been asked before..
if you dont interrupt you just get a ted talk.. this is a podcast ;)
No I don't think so. Dario was obviously stuttering at some points and couldn't quite articulate and didn't have the answer to the question. Nicolai cut him off so that they could move on to the next topic.
I'm so glad someone else said this. He consistenly interrupts him, it's a really strange interview tactic. You would get more from your interview subjects if you give them more time to think and explain themselves... If he stutters let him stutter. We want to see and hear how they approach the logic to answering these (mostly good and difficult) questions. You can't talk about global issues so quickly and clearly, it's supposed to be complicated, and I'd rather these conversations were given more space to breathe.
Yeah he starts his questions as Dario is finishing up his thoughts. Very annoying trait, probably useful in business because it puts your interlocuter on the backfoot, but its off-putting in an interview.
It's his style with everyone.
We don’t even know where out thoughts are coming from
Imagine being a billion dollar company and then wasting half of your money on preventing AI from doing something they've never been proven to do.
They are building a product for enterprise customers like Amazon and other major retailers. Ethics aside, its pretty important from a business standpoint that it doesnt tell you to off yourself or how to build a bomb.
Generally a pretty good and interesting interview but I was surprised by some the rudeness. Like asking if he had any friends growing up and saying he never grow up cause he plays video games
“… yeah, I mean, you know, ummm, like you know, yeah, you know like, I mean like it is like hard to tell like” ;)))
All the time I wondered how long would be this podcast if any active AI model cut all the useless, repeated information this guy said. … 8 minutes?
Reading tbe comme ts was expecting a hideous interviewer... he was ok.
Stop cutting him off … painful to listen too
I think we are listening to a heavily edited version of this conversation. The pacing is accelerated in post, but the actual interview was more civil. Just my guess. Feels edited.
"..like.." "....ya know..." ... "..like.." "....ya know..." and many times "...like ya know..." Aarrrghhh!
15 minutes in… snore
snake oil
To bad this guy is such a statist 42:00
such a terrible interviewer
Awesome insights & solutions to society’s stresses.