60% of decision-makers say their org is ready to deploy AI agents; only 36% of the workers using AI agree. On overall confidence in their org's ability to use AI, the gap widens: 49% vs 23%.
Leaders think their AI transformation is much further along than people using AI everyday inside companies. It’s almost like leaders and workers are describing two different companies.
@NotionHQ Hey, I wanted to apply for an intern role. I can share my AI portfolio if required. The thing is I am based out of India. Any chances I could still help in some way?
3. Training → combine both and predict the next word.
LLMs are foundation models. GPT, Claude, Gemini. All of them start this way. Then, they transition into Multimodals, but that is a story for another day (tomorrow?)
Spent quite some time learning about LLMs today. Turns out all they do is pretty much predict what segment is going to come up next. Here is what I learned in simple words. A thread:
Three things go into building one:
1. Data → millions of text examples to learn from
2. Architecture → the internal blueprint (for LLMs, this is usually a Transformer). It maps how every segment of text relates to every other segment.
The final one is the most crucial.
@Dhruvam987 Haha, "basically" is doing a lot of work there.
Its DISTRIBUTION is its moat. (the storage part is commodified, isn't it?)
I mean, what "differentiating factor" can you provide for "basically" storing code in folders that Microsoft can't copy?
(3/3) As a student, it screams to me that I should FOCUS MORE on concepts like System Design and DBMS than something like an actual programming language and the nitty gritties of it.
HMU if you sight a flaw.
(1/3) Read about Augmented Intelligence today. It basically says how machines should actually provide probabilistic output (analyses) about what should be done and humans should build upon them and judge. And then, the machine can execute it.
(2/3) However, how I extrapolate this is as: 2 tech leads (1 handles AI, the other handles overall judgement) instead of 10 engineers + 1 project manager.
@IBMwatsonx is actually a hidden gem. Got access to @IBM Cloud for 395 days. Let me see if I can build something. Orchestrate has my hopes up to be honest, got access to some of the cooler OSS Models.