jacobr1

11 ms·

151.
by jacobr1 2y ago
Is not the answer here the same? Build the trust in the AI systems. Benchmarks help and are a start. Consistent results over time are another measure. We'll learn over the next few years which models, or which companies developing mode
152.
by jacobr1 2y ago
The narratives around how such cultural traits evolve is not clear, and I agree are unlikely to be the post-hoc rationalization of "Great Men" setting them up. Yet there do seem to be some traits that enable more "success&quo
153.
by jacobr1 2y ago
Like the old (apocryphal?) saying from Ford that if they asked the customers what they want, they would have requested a faster horse. You can find similar statements people have found from old "man on the street" TV interviews ab
154.
by jacobr1 2y ago
> limitations of the person thinking of them, but not actual limitations of the medium of which work is done. I think we are agreeing here. That was the point of my examples, demonstrating limited thinking from the execs demanding RTO. I
155.
by jacobr1 2y ago
Semantic search without LLMs is already making a dent. It still gives traditional results that need to be human processed, but you can get "better" search results. And with that there is a body work on "groundedness" tha
156.
by jacobr1 2y ago
A support ticket is a good middle ground. This is probably the area of most robust enterprise deployment. Synthesizing knowledge to produce a draft reply with some logic either to automatically send it or have human review. There are both s
157.
by jacobr1 2y ago
Which is a good example, because accuracy can be improved significantly with even minor human guidance in task like unit tests. Human augmentation is extremely valuable.
158.
by jacobr1 2y ago
We aren't talking about skilled knowledge work in Silicon Valley campuses. We are talking about work that might already have been outsourced so some cube-farm in the Philippines. Our routine office work that probably could already have
159.
by jacobr1 2y ago
This is why we are only at the start of exploring the solution space. What applications don't require 100% accuracy? What tooling can we build that enables a human in the loop to choose between options? What options do we have to bette
160.
by jacobr1 2y ago
yep, or even just a text model ChatGPT 01-preview gave me: User: what is the Math squiggle that looks like a cursive p? Assistant: The mathematical symbol you’re referring to is likely the Weierstrass \wp function symbol, which resembles a
161.
by jacobr1 2y ago
To be clear, the examples I gave aren't those of what I think effective management should, but rather an amalgamation of the though processes I've observed. > These examples all seem outdated on both sides, and not realistic to
162.
by jacobr1 2y ago
LinkedIn remains a reasonable substitute for emailing around PDFs of resumes/CVs. Any usage beyond that, for "networking" or "social sharing" is terrible.
163.
by jacobr1 2y ago
>What I find entirely humorous about this is its executives that want hybrid / RTO by a large margin, and comparatively few employees want hybrid / RTO and prefer working from home. Converting an in-person culture to a remote c
164.
by jacobr1 2y ago
subleasing is an option, though demand is down across the board
165.
by jacobr1 2y ago
People get promoted for improving key metrics. Like Revenue or Retention or Installs or whatever. Metrics can be abused and gamed ... but you can do a better job aligning them to improvement than the de-facto metrics of "number buzzwor
166.
by jacobr1 2y ago
Could you explicitly train a set of embeddings that performed that step in the process? For example which computing the loss, you compare the difference against the normalized text rather than the original. Or alternatively do this as a fin
167.
by jacobr1 2y ago
It doesn't diminish - but I do find it interesting how it influences. Realism became less important, less interesting, though still valued to a lesser degree, with the ubiquity of photography. Where will human creativity move towards w
168.
by jacobr1 2y ago
Not sure that is the case. The base science and recommended practices seems pretty much as a solid as any other speciality (imperfect as they all are). It really is the case that cavities are bad. We understand what causes them, how they gr
169.
by jacobr1 2y ago
I think this presents an approach: move closer to the search engine model. In both academic writing an news journalism, factual claims try to cite their sources. And you can see this is an approach that some of the chat systems have directi
170.
by jacobr1 2y ago
In a microservice world obviously you'd have a feature-flag service. But you still have a build/buy consideration.
171.
by jacobr1 2y ago
That may be so, but the denigrate (and common case as this thread suggests) is to expand the notion to any offensive speech that is disliked by the offended person. That is much more subjective and hard to define. The fact that we have some
172.
by jacobr1 2y ago
>Is there a reason if it was unlocked you would immediately switch? If there was a better price or promotional discount available from a competitor?
173.
by jacobr1 2y ago
Also you can hire people with 20+ years of experience that also can't code (where people claim to be a software engineer). FizzBuzz was a real filter for a while. It has amazed me how some people where able to slide by in larger organi
174.
by jacobr1 2y ago
And that is one of the best ways to hire new grads. Take the best of the crop of interns you've had.
175.
by jacobr1 2y ago
One classic approach is to over-hire and weed out. I find some form of this de-facto happens anyway, so managing more explicitly has some benefits.
176.
by jacobr1 2y ago
It is all about variance. Higher variance activities have both more potential upside and more potential downside. Risk management is all about figuring out how to improve average upside. In investment, that probably means placing a variety
177.
by jacobr1 2y ago
denotation and connotation. They have the same literal meaning (denotation) but certainly not the same implication and tone (connotation). Choosing to eat a turkey sandwich vs roast beef is an opportunity cost. I'm forgoing the roast b
178.
by jacobr1 2y ago
While we can compare tail-risks (some lower probability risks and many orders of magnitude more likely than others) I think most people, intuitively, use something like a 1 in a 100 risk of a death (over a lifetime) or 1 in 10,000 acute cha
179.
by jacobr1 2y ago
Where is Claude more censored than ChatGPT?
180.
by jacobr1 2y ago
Also Anthropic, even though it has the same conflict of interest as OpenAI, seems to be addressing safety concerns more earnestly. They are listening to their internal team and innovating in the space (especially in a practical sense, not j

More ›