I’ll throw in another anecdata point (more than one as I have multiple children): I found them to be mostly logistical burden for the first few months he is writing about, but my love fully developed over the years. Don’t give up hope if you’re not feeling it instantly!
They seem to really prioritize “just works”. I am currently reviewing a PR that assembles some code as a string, using a CSV with carefully named columns. It does, in fact , work on the example CSV. I would invite this team to approve and support that for me over the next 5 years.
I've been on both sides of the table for both full-time and consultants. The first rule of humans is that they vary widely within categories (even if there are statistically significant differences in population average), so I urge people to take all of the anecdotes on this page with many grains of salt.
That said, here are some of my own, condensed across experiences:
- An executive wanted to use ML, couldn't get enough political capital to build a full-time team (and didn't have the knowledge to interview candidates anyway). A very small consulting firm, sourced from high personal trust networks, got an MVP going. That led to a FT team getting built. Code was a mess but started the flywheel and after several years everyone agrees it was a great ROI
- A large company had a tiny team using a technology that would be standard for FAANG but they had no support or training. Zero progress for about a year. Not important enough to throw huge money at but that link in the chain was becoming a bottleneck for the core product. A consultant came in explicitly as a multi-week trainer, and but the team just never got up to speed. Consultant's code examples grew until they were just direct bugfixes, and though nobody was happy, at least the crisis was resolved.
- A product finally reaches product market fit. Executives decide to hire only after the team is drowning in tech debt. But it has to be full time, and it has to be junior. It takes months, then requires training. Tech debt is twice as bad as the original crisis level when the new hire becomes value positive. For some bizarre reason he leaves after exactly 2 years.
Many consultants have a branding problem: Their biggest victories are fixing a disaster that their clients would rather not talk about publicly. Hard to put on a billboard, if you ever want work from that client again!
1. What other course of study are you confident would be better given an AI future? If there's a service sector job that you feel really called to, I guess you could shadow someone for a few days to see if you'd really like it?
2. Having spent a few years managing business dashboards for users, less than 25% ever routinely used the "user friendly" functionality we built to do semi-custom analysis. We needed 4 full time analytics engineers to spend at least half their time answering ad hoc questions that could have been self-served, despite an explicit goal of democratizing data. All that is to say; don't over estimate how quickly this will be taken up, even if it could technically do XYZ task (eventually, best-of-10) if prompted properly.
3. I don't know where you live, but I've spent most of my career 'competing' with developers in India who are paid 33-50% as much. They're literally teammates, it's not a hypothetical thing. And they've never stopped hiring in the US. I haven't been in the room for those decisions and don't want to open that can of worms here, but suffice to say it's not so simple as "cheaper per LoC wins"
> I suspect all the aid and help over the decades has really just prolonged things
Can you be more specific about the counterfactual? I’m assuming you’re not imagining that the country’s population would have just fallen to 0 absent aid, but I can’t figure out how you’re so sure things would have been better run
I didn’t write the comment you replied to but often what happens is that politicians in Wealthy Country send aid to Poor Country so they can say they did something. The aid often props up the Baddies in Poor Country and doesn’t reach the people who need it most.
The Dictator’s Handbook covers this topic I think.
Treating “Sudan” as an actual entity to be helped. Consider it has failed as a political concept (the state of Sudan) and allow the people that inhabit it figure out what the borders and political entities actually are. By delivering aid to Sudan and treating it like a real entity you only prolong suffering in the region as factions fight over control of it.
I think that’s correct – even at a “high” cost (relative to what? A random SaaS app or an hour of a moderately competent Full Stack Dev?) the ROI will already be there for some projects, and as prices naturally improve a larger and larger portion of projects will make sense while we also build economies of scale with inference infrastructure.
You can just use Gopher, you don’t need Google for this.
I’m being flippant, obviously but my point is that every now and then a new technology comes along that supplants and older one because it offers a meaningful improvement over the last. AI search vs traditional web search is definitely one of those.
My multiple experiences with ChatGPT hallucination, even at minimum temperature, leaves me preferring a web search answer. Just last week, jq hallucinated three different jq syntaxes that failed to parse.
That being said, I had been using web search for several years looking for the Get Smart episode I had seen as a kid in the 1980s where the enemy agent explains that he can't give up his suicide ring because "That's how it works, if I take it off, my wife will kill me." Google Bard found the episode on the first try.
So, I tend to fall back to LLMs only when search engines fail me.
You’re not thinking through the whole experience. The search itself is just as easy, sure. You can type the same query into both systems. The difference is in the results. Traditional web search will then return a set of links which may or may not answer you query, it’s up you you to parse them and figure out if they have the content you need or if you need to rephrase your query and try again. AI on the other hand will parse the results for you and return the answer to your prompt directly. That’s a meaningful improvement.
> AI on the other hand will parse the results for you and return the answer to your prompt directly.
Yep! And that's going to save you _one whole step_ when you're posting meaningless follow ups in forums filled with technical users who would have had no problem getting the information themselves.
> That’s a meaningful improvement.
It's a marginal improvement to the least meaningful activity that occurs on a daily basis. So, it's totally worth the trillions that have been invested in it. :|
which is exactly what Perplexity did. I am saying this out loud because it'll be a few years before people wrap their mind around the circus over there. not because its awesome.
1) pay google scraper api (~1000 searches/$1)
2) run the crappiest llm people won't notice is crap
People might not like the AI plug, but personally this is the one use-case I find AI the most helpful to me. The ability to query general knowledge, with natural language, often not knowing what the hell I'm talking about. And the results don't have advertisements, which I also really appreciate.
I am so impressed with the work of these bloggers. The lawsuit from Francesca Gino they’re dealing with in the Harvard Business School case, and death threats in this one. It takes courage like this to make a systemic change.
Look, I basically agree with the goal and am trying myself but: Try watching the Little League World Series on YouTube twice weekly for 4 years and you will have 1% of an understanding of why this is hard.
Its addiction and FOMO, you form connections. Even if you never talk to folk, you have an instant connection if the topic is to arise in face.
However if you previously and stopped watching and the conversation comes up you don't have have the up to date information to connect in to the conversation.
So because your part of the community by whatever feed you feel part of the community. The fear of not being part, subconsciously is what drives you to keep up the habit.
Same goes for all, communities, fandoms, cults; toxic or not.
