marckraw

[Mitchell Hashimoto @mitchellh]

The "whiteboard defense:" I should be able to pull you aside at any moment and ask you to explain any customer-facing system you've shipped. You should be able to clearly explain how it works and defend the decisions you made. This is my benchmark for responsible AI usage.

I don't expect line-level familiarity with the code. I don't care if you remember the exact function name or implementation detail. You may not even know it. I don't care.

But if I ask "why did you do X instead of Y?", "what happens if this actor behaves maliciously?", "what data structure did you use here and why?", or "where does this fail?" you should be able to answer confidently.

For PoCs, demos, experiments, whatever: I don't care. Generate 100% of it and understand none of it. Speed over quality every time in those specific scenarios.

But if you're shipping customer-facing work, you can't be shipping things you don't understand at a high level.

[kitze 🛠️ tinkerer.club @thekitze]

we should care about the code
we should read the code (a bit)

we shouldn't do this by opening file by file
we shouldn't go and fix it manually

we need new tools.

[marckraw @marckraw]

@andrzejdragan to jest absolutnie genialne <3

[marckraw @marckraw]

@pie6k any chance of cyber monday extension to the deal ? :D

[Jan Leike @janleike]

Yesterday was my last day as head of alignment, superalignment lead, and executive @OpenAI.

[Adam @_overment]

A few words about real-time interaction with LLM and the road ahead.

- Real-time interactions are expected by end users, yet response quality is the key to success.

- A valuable response requires understanding the message in the broader context of the conversation, considering long-term memory, environmental data, and external sources. This process can be achieved by an LLM, relying on predefined rules and application programming logic.

- Thus, a single user message may involve multiple queries to the LLM. This process takes time and incurs costs, but significantly increases the value of the final response.

- Models like GPT-4o understand content thoroughly and can transform it in various ways. They create action plans, implement them step-by-step, use tools, and assess progress.

- It's possible to create a general procedure for an LLM to react to received data and decide independently, but human supervision remains crucial.

- Models like GPT-4o can transform content based on specific rules, showing a deep understanding. They can plan actions, decide on next steps, react to results, predict future actions, and analyze the emotional state of the conversational partner to adjust their responses!

- Despite advancements in GPT-4o, parameters for reasoning, speed, and cost still struggle with complex logic, but significant improvements are pushing the boundaries of what's possible.

- It remains true that LLMs perform best when highly personalized and tailored to specific processes, e.g., operating within an organization or directly for us.

- The path to "production" requires further increasing reasoning capabilities, speed, stability, and significant cost reduction. This also includes the need to develop tools, ecosystems, and knowledge about the current capabilities and limitations of the models. It's still very early.

To give all this some context:

- OpenAI integration costs, currently used solely by me, reach approximately $500 per month.

- Execution time for a simple command ranges from 7 to 25 seconds.

- Execution time for a more complex command, involving several steps, quickly exceeds 1-2 minutes for GPT-4o

- About 10-15% of actions fail, which is acceptable internally since the rest still adds value. However, for production, this is unacceptable for most users and business itself

- Despite the above points, we're still talking about huge time and cost savings, often enabling actions I wouldn't normally have the resources for.

More coming soon... 🙃 Next time, I'll show a few practical examples from my daily workflow. Stay tuned!

[marckraw @marckraw]

@K_Stanowski Izak, do esportu skoro wspolpraca z xkom :)

[marckraw @marckraw]

@wina_Mazurka Jarek "Pasha" @paszaBiceps potrzebujemy esportu! ;)

[Enzo Avigo @0zne]

“AI will destroy the world”

Meantime AI:

[Greg Brockman @gdb]

After learning today’s news, this is the message I sent to the OpenAI team:

[Sam Altman @sama]

i loved my time at openai. it was transformative for me personally, and hopefully the world a little bit. most of all i loved working with such talented people.

will have more to say about what’s next later.

🫡

[Krzysztof Stanowski @K_Stanowski]

Proszę o RT.

Natalia Janoszek myśli, że może zakneblować mi usta i zabronić ujawniania prawdy na jej temat. Otóż nie. W ten czy inny sposób wkrótce ukaże się materiał na jej temat, bo po to pojechałem do Mumbaju i nie tylko. Po fakty. I czy jej się to podoba, czy nie, ja te fakty ujawnię. Kiedy, droga Natalio, nazwałem cię patologiczną oszustką nie wiedziałem jeszcze, że to zbyt pieszczotliwe określenie. Jesteś polską Anną Delvey vel Sorokin.

Stop cenzurze prewencyjnej.

[marckraw @marckraw]

@NicEstrada3 @ThePrimeagen yeah same for me :p

[Matt Pocock @mattpocockuk]

TypeScript annoys me sometimes.

So, I decided to do something about it...

[marckraw @marckraw]

@devagrawal09 @charafmrah and that is their decision i think, and as you can see there is a split in thinking about this situation, so its not black and white what is "the way". I've watched the reaction and the video, and I think there will be no reason for ppl to watch both because reaction unweils all

[marckraw @marckraw]

@honeypotio @t3dotgg @t3dotgg very mature of you, hiding and blocking people who dont agree with your one and only view of the world

[marckraw @marckraw]

@t3dotgg @honeypotio gj @t3dotgg very mature

[marckraw @marckraw]

@t3dotgg @honeypotio Yeah the response like that is giving no respect to how they would like to distribute their hard work. "totally missing the way this generation of content works". and who the f are you to dictate them anything... weird man, weird

[marckraw @marckraw]

someone didnt like listening to other opinions :) such a sad story of amazing developer, who started to be I Know Everything Better. Note to anyone to be humble and respect others and others opinion to not became sad little man.

[marckraw @marckraw]

@devagrawal09 @charafmrah thats what you think, so it could be totally opposite of what they think, and they have right to have different way of looking at that