Warranty data is a great example of where LLMs have evolved bureaucratic data overhead. What most people do not know is because of US federal TREAD regulation Automotive companies (If they want to land and look at warranty data) need to review all warranty claims, document, and detect any safety related issues and issue recalls all with an strong auditability requirement. This problem generates huge data and operations overhead, Companies need to either hire 10's if not hundreds of individuals to inspect claims or come up with automation to make this process easier.
Over the past couple of years people have made attempts with NLP (lets say standard ML workflows) but NLP and word temperature scores are hard to integrate into a reliable data pipeline much less a operational review workflow.
Enter LLM's, the world is a data gurus oyster for building an detection system on warranty claims. Passing data to Prompted LLM's means capturing and classifying records becomes significantly easier, and these data applications can flow into more normal analytic work streams.
I am not a fan of their initial "Global Income Distribution" curve. if you take the actual data at the bottom of the article and plot it; it does not make anything the resembles a standard distribution as portrayed. It could be an infographic, it could be different axis, who knows, but portraying a standard distribution is wrong if you have an outlying skew in your distribution. Everything under $40 is a standard distribution, but above $40 represents the same volume of people as the average skewing any sort of plotting.
Well written. Coming out of college years ago Gartner was a whole section of review during my business courses. Working with Data for years now I have become hyper sensitive to this keyword grift; Big data, Data lake, Datalakehouse, realtime-analytics, no-code, data model, data schema...etc. People lean so hard on certain words as if they mean they are doing something different or unique. You work in one product in your company, then you bring someone who has experience in another product and they remark "But product X cannot do XYZGrift" but it can, people hang on these keywords as though they are platform actions or enablement that exist only there.
Rambling, but to get to the point, AI in general will strip this SEO/Marketing/Boomer catch phrasing, and build the common language which I appreciate greatly. I can go to ChatGPT or Claude and ask it I want to Foo this Bar with these filters, doesn't matter if its SQL, Python, Unix, Alteryx, Tableau... whatever, it digest the request without the fluff and responds commonly.
To stack on this info hunting or product research with AI is also typically less full of fluff for me. I don't have to deal with a sales engineer saying how wonderful their ML product is when I know its garbage immediately, I can just move on and assess the rest of the product.
The only value I can still see in Gartner is their customer survey information, but I am sure someone or somehow AI will scrape the forum post for all these products and weight the products community feedback about its product.
>"...we are looking at it, we are looking at him, and he better watch out, because that's not a protected speech. That is threatening the lives of our law enforcement officers throughout this country."'
wild statement from the person who went to law school, but threw out everything they learned.
I see little to no difference between this, Waze, helmet* taps, or flashing your high beams to other cars when passing the cops. That topic in general has been in court multiple times, and every time the ruling was in favor of it being considered freedom of speech.
I was told a story when I was younger (take it with a grain of salt I cannot find anything to corroborate it). The British embassy use to offer (maybe still does) "Tropical" pay for individuals stationed in temperate climates. Washington D.C. was considered a tropical location for years because of the notorious swampy and muggy conditions experienced in the warmer seasons. Stationed diplomats knew of this hazard/tropic pay and wanted to keep it, and when leaders would come to visit they would exasperate the conditions by turning off the AC. One year some time in the 80's they forgot to turn off the AC during a prime minister visit, and at that point the tropical pay was revoked.
Also fun fact they also have a Pub in the basement of the embassy.
Absolutely, in the past some tuning companies had no issue recommending removing emissions equipment, but then standards and laws changed and most if not all companies complied. Banks power is a good example of a diesel tuner who does not advise on removing EGRs.
In the area of emissions management aftermarkets have found symbiosis. Not only has there become aftermarket manufactures of higher performing both in emissions and performance catalytic converters (G-Sport by GESi). But alot of tuners acknowledge that engine performance can be improved with a well catalyzed exhaust system.
Look at Holley aka Dinan aka APR. They are a publicly traded tuning company. The aftermarket tuning they offer meets emissions standards while still offering better performance than OEM. So yes they are not "legally" required, they still do it to meet state and national laws and to avoid toe stepping.
Yes the little guy tuners don't really care about NOX emissions guidelines mostly because they have no way of measuring the difference their tunes may produce. At the end of the day most OE emissions systems will catch these emissions, and pretty much all aftermarket tuners do not endorse removing emissions systems. to say a tune, or aftermarket parts result in worse emissions is not really the truth. to say the end user removing their cats or emissions systems results in worse emissions is true.
This article is covering two different topics and trying to make it seem like one thing that it is not...
1. GM is using an ML model for their "torque management" which is a fancy of of saying a linear feel pedal.
2. This new generation of ecu has more encryption... every new generation of GM ecu has more lockouts.
The author alludes to how tuner will not be able to beat GM's torque mapping controls with aftermarket tuning. Sure... but often times turners are not targeting the drivablity mapping of an OEM tune, they are targeting fuel, ignition, and or boost mapping to compensate for better fuels, more VE (turbos), or other power adders.
TBH on flagship sports cars, we are on the knifes edge of optimizations for most platforms; most aftermarket solutions are now just lop-siding the maximized "any condition" performance OE's seek for simply more power. Power that typically will sacrifice either low end, drivability, and or reliability. The sweet spot for performance tuning now a days exist in the middle range of vehicles for most manufactures where engines are not focused on their ultimate tuning potential VS reliability.
This all being said, these torque management strategies are nothing new, GM is just using fancy math blocks within their ecu that can account for more inputs and a higher resolution. Modern standalone ecus like Emtron and Motec utilize these types of torque strategies to better pair with modern high end transmissions like the 8hp90 and DL800. These transmissions need to communicate with the ECU to ensure power delivery from the engine works with shifting performance and clutch engagement.
This is a double edge sword problem, but I think what people are glazing over with the compute power topic is power efficiency. One thing I struggle with home labing old gaming equipment is the consideration to the power efficiency of new hardware. Hardly a valid comparison, but I can choose to recycle my Ryzen 1700x with a 2080ti for a media server that will probably consume a few hundred watts, or I can get a M1 that sips power. The double edge sword part is that Ryzen system becomes considerably more power efficient running proxmox or ubuntu server vs a windows client. We as a society choose our niche we want to leverage and it swings with and like economics, strapped for cash, choose to build more efficient code; no limits, buy the horsepower to meet the needs.
I'm not a fan of this trend either. My suspicion is this change is to increase scrolling to pump more ad space; it makes sense from a business standpoint. But this combined with the Algo changes makes it hard to keep coming back looking for new content VS just consuming the people/content I know and enjoy.
As others have pointed out VIN as your primary level of detail is fairly flawed. You should also be filtering on vehicle registrations. A possible scenario here is someone owns their car, gets into an accident, repairs it or sells it off (insurance sell off). Someone else buys it gets difference insurance, and has an accident because its a new car to them. You would never know this vehicle story without integrating the registration history or at least buy and sell dates to know it has exchanged hands.
Yes Fraud is tricky, but a VIN does not equal a person (PII) committing the behavior, and these poor association attempts leave innocent people screwed by insurance companies. What makes this bad analysis annoying, is the constant caveats, if you know VIN associations don't prove fraud then maybe don't build some sort of risk scoring on it, and then try an sell it.
"Hadoop’s meteoric rise led many organizations to implement it without understanding its complexities, often resulting in underutilized clusters or over-engineered architectures. Iceberg is walking a similar path."
This pain is too real, and too close to home. I've seen this outcome turn the entire business off of consuming their data via hadoop because it turns into a wasteland of delayed deliveries, broken datasets, op's teams who cannot scale, and architects overselling too robust designs.
I've tried to scale down hadoop to the business user with visual etl tools like Alteryx, but there again compatibility between Alteryx and hadoop suck via ODBC connectors. I came from an AWS based stack into a poorly leapfrogged data stack and it's hard not to pull my hair out between the business struggling to use it and infra + op's not keeping up. Now these teams want to push to iceburg or big query while ignoring the mountains of tech debt they have created.
Don't get me wrong Hadoop isn't a bad idea, its just complex and a time suck, and unless you have time to dedicate to properly deploy these solutions which most business do not, your implementation will suffer, your business will suffer.
"While the parallels to Hadoop are striking, we also have the opportunity to avoid its pitfalls." no one in IT learns from their failures unless they are writing the checks, most will flip before they feel the pain.
AI focused new BI products are a wash, your focus should be a new BI product with maybe an AI integration (Not the other way around), or an AI integration to an existing BI product. The BI tool Kanaries / PyGWalker is a good example of where python based BI tooling should be going, replacing big corp BI tools to get CDO/CIO's to go the direction of abandoning Tableau/PBI for lower infra and platfrom cost, open source tooling & integration.
I am not saying its a bad idea, but your net is wide, very wide, you are trying to conquer two separate area. Both areas need innovation in the data space, but I think the AI rush is dieing down. Claude and GPT offer good support for PowerBI and Tableau, but there are no extension integrations. This is where BI AI should be, AI's integrated into Tableau workbooks editing the core XML of the workbooks, offering translation tooling to convert Prep to Alteryx, Tableau to PBI...etc.
It's also hard sell if your core platform doesn't cover SSO, Encryption, DB integrations, Self Deployment, RBAC's, PenTest, Corps looks for these things immediately, and BI tools are purely a Corp thing not a Consumer thing.
This is so often missed by people in the auto industry, I am glad he is saying something. But this is not just limited to vehicle components within hardware of cars; this also will also often apply to the companies entire tech stack. By this I mean vehicle ordering systems, parts management systems, dealer systems... etc.
The auto industry for years has had a "Vendor Solution" problem which has created insurmountable software and hardware tech debt or dependencies. Some manufactures are using systems dating back to the 90's in production. Companies will deploy newer better DB's and never migrate off the old solutions and this leap frog happens multiple times so you have 3 or 4 "Generations" old infra. Most of this is the result of farming out or contracting vendors to build a system, deploy a system, and it just stays. After which all of the knowledge vanishes about how it was built, then business owners want updates and no one know how to, so its farmed out to figure out how to fix it... this cost $$$ and time, there is scope creep, and the cycle repeats.
This is starting to change, but Farley's right in describing this is a HUGE problem the public doesn't really see beyond just the vehicle systems. If you are unsure if this is true, this is why the CDK Global hack essentially froze all vehicle sales and dealer pay, CDK is a huge DMS (Dealer Management Systems) provider that dealers pay to do their parts and sales systems.
It was profitable to do this in the 90's, 2000's, and 2010's due to globalization and smaller digital foot printing, but that has all changed with IOT, Smarter sensors and controllers, and consumer & CEO expectations of product advancements.
Almost all modern cars have a way of providing or grabbing location data, however most manufactures do not "Spy" on your car by default, this would violate CCPA, colorados privacy act, GDPR... ETC. The users need to opt-in to telematics data. For example in Hyundai case when you create a "Blue link" account and accept their terms of service you are connecting whatever vehicle you have verified on your account to their telematics system, and subsequently opting in to tracking.
Manufactures like VW/Audi place an opt out within the vehicle itself so if you opt out of telematics in the vehicle you are in a full privacy mode and the manufacture cannot get the data or override this request. This covers the scenario if other "Users" of the vehicle are driving and would choose to opt out outside of the main users/owner.
So some bake it into your app registration and signup, and some leave it in the vehicle. The gist is you can opt out, and if the manufacturer does not respect that you have grounds to sue, Currently there is a lawsuit against GM/Caddy because a user did not opt-in to Usage Based Insurance, but their information was captured and brokered blocking them from acquiring new insurance.
Over the past couple of years people have made attempts with NLP (lets say standard ML workflows) but NLP and word temperature scores are hard to integrate into a reliable data pipeline much less a operational review workflow.
Enter LLM's, the world is a data gurus oyster for building an detection system on warranty claims. Passing data to Prompted LLM's means capturing and classifying records becomes significantly easier, and these data applications can flow into more normal analytic work streams.