Benedict Evans: AI Eats the World | Slush 2024

[@JesperAndersen]

Great video - too bad they gave you such a bad microphone.

[@KingTine724]

Yeah 😂

[@AIForHumansShow]

love this but good lord the background noise

[@akrs]

Ironically, AI is pretty good at removing such noise.

[@its_ot]

I started this video, got distracted, and caught the last metaphor about elevators and was like "when was the last time I was in an elevator that wasn't an Otis?"

[@miraculixxs]

Schindler

[@idkwhymadethisaccount6000]

mitsubishi, kone, thyssen krup etc etc

[@Rodrigodemetrio]

great content

[@bureau_works]

Great session

[@GrindThisGame]

AI is easy, AV is hard :P

[@deeplearningpartnership]

OpenAI started in 2015, so it took nine years to get to a $150billion valuation (not 18 months as he claimed).

[@benedictevans7084]

Microsoft didn't' take 9 (or 7) years to start selling something. And I wasn't doing a DCF, just illustrating the relative speed ;)

[@deeplearningpartnership]

@@benedictevans7084 Yes it's happening fast alright, but no need to make false comparisons. And AI has been around since the 1950's if you really want to be up front and honest. But I get it, hype sells.

[@benedictevans7084]

@@deeplearningpartnership 'AI' is a concept, not a technology. It's no different to 'automation'. That doesn't mean that LLMs are not a hugely important and powerful new technology.

[@david.walters]

OpenAI launched its first product, the GPT-2 language model, to the public in February 2019 as a research release. However, the full commercial launch came with GPT-3 in June 2020, when OpenAI introduced its API through a private beta program.

[@benedictbleimschein4526]

Awesome content, but why do i have to watch this in 320p? No budget for proper cameras and audio?

[@ausderek]

They spent all the money on the lasers 🙂

[@shyft09]

in this case it's probably your crappy network (at the moment I see up to 1080p)

[@henryebube3576]

I totally agree

[@BassForever44]

Extremely interesting. Bit ironic the tech theme accompanied by such horrid audio mixing. Fire the sound engineer!

[@alveek]

Watched a half of the video and gave up What is he trying to say?

[@ianr4222]

Why does he keep referring to ChatGPT and OpenAI? I use Generative AI most days in my work as a Software Engineer. I have access to OpenAI ChatGPT-4o, why would I use it instead of the much better model Anthropic Claude.Sonnet 3.5, which I also have access to.

[@kkb92-96x]

What is wrong with AI being an automation tool? I don’t understand why the speaker is denigrating AI as just a software and all it does is automation. AI is software, no one is claiming it to be sentient. It is still very early stages of Gen AI, as he himself indicates businesses have still not fully embraced cloud, this will take time. There are tools now to develop agentic applications that can make use of AI. It is revolutionary in the sense that you don’t have to rewrite the code for every little thing. I see all of the SaaS software will be the first to embrace AI and provide value added services. Salesforce is already doing it, so are many HR related SaaS services.

[@loganholdaway769]

I don't feel like he said a lot. Interesting point that all other major tech rollouts have had a specific use case in mind, but not here. Personally, I think it's great that we are at a point where we get to be creative with such a massive disruption. I'm disappointed that he said AI is any tech in the early stage of adoption. There are similarities, yes, and not everything that carries the AI label is AI, which confuses the issue. But AI is more than just a new automated something. AI involves complex higher level thinking with natural language. The elevator computer isn't AI.

[@GrowStackAi]

At this point, AI is smarter than all of us combined 🔥

[@cakep4271]

Current ai is all basically Google search + good summarization. A very useful and flexible tool indeed, but not great at dynamically understanding unique systems, of which our entire society is made of.

[@kkb92-96x]

I think your comparison to search maybe a good way looking at it, but much more efficient than current Google search. I say this from the examples I have read like finding some obscure information in a scientific journal that everyone missed or finding some chemical reaction that researchers never found. Microsoft posted a university of Washington use case of some metallurgical research where it used to take 5 years was able to complete in 1 year. These are still search but effective. At the same time, these effective search when orchestrated with series of agents can disrupt the way we do things. These are still very early stages.

[@bencium-agentic-ai]

This is the dumbest approach ever heard so far