Has it delivered anything anywhere?
21 replies (most recent on top)
@222 you can use more things than just copilot if you actually know what you’re doing. I’d argue that even if you got to use ChatGPT and perplexity plus copilot, it would still su-k. Those other things give you slightly better user experience, but if you’re actually building real AI solutions, you’re going custom stuff, using open-source frameworks to design and build what you actually need anyway.
Whether it’s Gemini, Claude or GPT - they are all fairly locked and inflexible for you to fine tune or actually design things that fit your exact need.
And I would say that if you know what you’re doing as a developer, you do have the cloud services, resources, and open source things to do things such that “AI wouldn’t su-k at Chevron”. There are quite a few people who do great work… and then there are those who are just asking to buy licenses because they want a nice user experience without actually understand what they actually need to do.
Does CVX let you use any other AI tools other than MS Co-Pilot? I remember they let us use ChatGPT for like a year after it was introduced then blocked it along with Gemini, Perplexity, Claude etc....
If all you get to use is MS Co-Pilot than AI su-ks at Chevron.
In the near future, my vision is that AI should be able to replace everyone in management, allowing the rest of us the freedom to get on with the task of finding oil and gas without all the constant disruptions.
If you don't think the reimman hypothesis has been proven already just try to disprove it. You can't
I worked on a project where a guy used AI to write the final report and it was horrible. The guy writing the report did not even know the report was not accurate because he did not know what he was doing. The old saying is true, Garbage in = Garbage out
OpenAI definitely does not have their data in shape. Databricks’ biggest customer is now OpenAI but that’s only a recent thing.
They were able to do some pretty great things without having all of their data in great shape. I think that was the point.
No wonder there’s no adoption? Adoption of what? What are you even talking about?
There’s certainly adoption of m365 things, and stuff that doesn’t “scale” - why do you need everything to scale? That’s the reason why the center and their braindead idea of scaling has caused no adoption. Because all you id--tic IT people think that GOM has anything to do with MCBU and stuff needs to scale when they do not.
Most of the adoption issues have to do with business workflows, not about data or technology.
they literally do. have you heard of scale ai or mercor? that’s who sells it to them. how about apache airflow? that’s how they move it all around when it’s not being slurped up for a training run
you could not have made a more ignorant comment if you tried.
you can make an argument for the fact that you can just upload a document and chat with it and that has value, but that’s an insecure thing to do since these companies then have a copy of the corporate data and it doesn’t scale to problems where your answer isn’t there.
i took it from the context of mentioning basic stats instead of say, training runs, that the idea there was classic algorithms like regression. what was your interpretation? do you have more riveting thoughts on how ai doesn’t need data ever? no wonder we aren’t adopting it.
you need data above a certain quality threshold available as near real time feeds to scale AI solutions
people like you making point solutions that don’t replicate and scale is exactly the problem. taking random teams pdfs and excels to make something is a proof of concept, not a deployment.
@gb actually, no
AI is a lot of math, statistics, and probability. It doesn’t need exact or perfect attributes
“Great data sources” as defined by IT consulting companies about data quality and information management are not really a requirement to do AI anymore. If they’re really good, sure, it helps.
But it’s a loser mentality from our old school IT data architects to say garbage in, garbage out, you can’t do AI without good data - when they don’t even know what that means and as usual, just parroting talking points with no actual understanding of our business.
AI is doing enough in your day to day such that a few years from now, you will look back and see that things have definitely changed. Just think of how work has changed even before 2020.
The real answer is no. It is the flavor of the month and is dependent on great data sources, of which there are none.
Some leaders are confusing automation with AI. Like it or not, “AI” has gotten their attention and buy in to finally do something.
SCM, HR, IT - the writing is on the wall. A ton of things will be automated, with or without AI. I know of a couple of projects going on behind the scenes that is even automating the processes that just recently moved to low cost areas. And quite honestly that’s how it should be. Leave those few workflow owners in the US, and don’t move jobs to low cost centers at all if you can just automate it with “AI”
"AI" can summarize text and provide a much better version of the Google search. That's basically it. Until AI generates some new piece of knowledge previously undiscovered by humanity, such as solving a Millennium problem like proving the Riemann hypothesis, I see nothing but marketing hype.
@bz they’re not cutting AI spend.
They’re cutting IT spend.
They moved AI out from under LC and are allowing JG to fund whatever he wants in AI.
AI is the buzzword but in cost cutting mode, it does nothing. Waste of breath when they think AI solves all of the problems. Look into what the problem really is. AI is not the answer to everything just a way to justify layoffs
when the technical leaders aren’t technical, what happens is that traditional MLop, i.e. model tuning to solve real problems like predictive equipment failure at scale, has its budget cut to make way for LLMs because “it is all ai”
smart technical leaders only buy into technology when it delivers proven value. there are very few use cases for gen ai in oil and gas. and they all were doable with OCR/ traditional NLP and data pipelines, just not as easily as now. for example, filling out contracts or being a search engine where engineers can find tables or other info they need from thousands of documents without wasting time manually hunting.
genai helps coding and that’s kind of it. and so clueless LT throws LLMs at every problem, gets no results, headlines come out saying 99% of ai pilots fail, then they cut all AI spend.
clueless.
@a1 Even better if AI can write end of year performance reviews.
AI can crunch through data and find things that a human could given enough time. I use AI to code, to compare, to create. Can it solve all the issues? no of course not.
AI = Actual Indians.
CVX = Delivery++
It depends on how you’re defining AI.
People are incorporating AI in how they do work. If you think work at Chevron is valuable, then sure, AI is doing things.
But if you’re defining AI as the old data science work where we had a really good crew, then I would say the value has been eroded thanks to LC. He will be also known as the guy who destroyed data science at Chevron
I wrote my PMP with copilot in 5 minutes.
What a time saver.