Friday, 13 October, 2023 in Portland, Oregon and it is grey and not warm.
Registration is now open for Hallway Track 002 Journalism, News, and Federated Social Networks. 12 spaces open today, the rest opening up on Monday.
These are the kinds of things that pop into my head at the intersection of "oh no, this would be horrible" and "horrible because they are on some level plausible".
Mostly they're fake headlines or quotes.
Also: those quarterly youth intelligence and trends reports from those late 90s/early 2000s research agencies that cayce pollard would do work for, but weekly, for $40/month, and for affluent/middle-class helicopter parents wanting to be best friends to their kids.
(Don't do this.)
(On preview, I've retitled this The Torment Nexus Section3; it was originally titled Needlessly Horrific Provocations)
OK, so Wired did a really interesting / horrifying story about how LLMs (automatic text generation engines and word/phrase-relationship engines) have upended the whole business of responding to RFPs4, or requests for proposal, which is how really big companies ("enterprises") buy software that their employees and anyone in their sphere of influence has to use, and is not consulted on. It's basically a great experience all around narrows eyes for people.
I promise to make no Star Trek jokes in this bit, but realize now that after having laid that particular Riker's Phaser on the conference room table, there's an increasingly large likelihood it will be set to kill.
The article is fantastic because it wonderfully illustrates how often the application of technology "solves problems" in potentially the easiest and most efficient way because it doesn't rely on anyone doing more work to do things properly.
In a way, RFPs make sense! They are like if lawyers, who generally are supposed to help decrease the likelihood and severity of bad things happening had a party with people who don't like to spend money and met their new best friends, over-controlled people who don't like to change things because change is terrifying.
Bad things have happened in the past when a company has tried to do things (e.g. it didn't get what it want, wasted money, was defrauded, the change was traumatic, the supplier disappeared, the supplier did something completely different, the thing was the wrong color, the thing was too many things, the thing was too few things, the thing didn't actually plug in to the other thing, the thing blew up, the thing blew up in the wrong way, etc. etc.) and the RFP is the resulting document that only ever gets longer and more complicated.
So. There is bullshit work on the RFP-generating end, because new specious requirements keep getting added. Like, I don't know, RFPs for large government systems like unemployment insurance end up including requirements like "you can only bid if you have made a large unemployment insurance system", which on the face of it makes sense, but on the other hand, restricts your field of bidders to a bunch of bidders that, let's say, have not proven to be very good.
You will of course have figured out that there is a bunch of bullshit work on the RFP-parsing and responding end. RFPs aren't standard, because that would be helpful, and it would be harder for them to be standardized. They are like Shitty Enterprise Job Postings: you see a job posting you want to apply for, you get to upload your resume or cv, but it turns out you still need to fill in a form anyway, the form requires different information, the form then goes into what appears to be /dev/null5 or an incomprehensible perl script turned into a batch of people, and then in 2023, some "AI" gets thrown at it that will then practically ignore or not "understand" the content of your resume in the first place.
So. I've been involved both in RFPs (trying to make them less shitty and more open to competition, and more likely to produce good results to the people who have to actually use the thing to get a thing done), and responding/parsing RFPs (in order to do work for people). I hate it. It takes forever. One submission I made, which also had to be hand-delivered by paper, and ended up being hand-delivered in a costco bulk box of hawaiian potato chips, including supporting documentation, was an eight hundred and twenty eight page PDF. And me and one person had to put it all together. (We won the work).
Anyway. Nobody likes responding to RFPs, and I reckon RFPs are at least 80% bullshit jobs6. Well, Twilio built an internal tool that: a) is fed an RFP and b) automatically generates a response. Note that I did not use "AI" in that sentence.
(This raises the amusing possibility of using that tool to generate statistically probable text of an RFP response in the manner of, say, a Ferengi7).
Twilio isn't the only one to do this, and you would be wise not to take me up on this bet:
The Wired article even points out the stupendous waste of time stupendous numbers of flops and watts (ha. ha.) are now being applied toward. These are direct quotes from the piece!
"RFPs with tedious questions asking for information such as a cloud service’s uptime or its support for multifactor authentication are an unavoidable stage in all sorts of sales processes, from software deals to light bulbs"
"[but] no one enjoys writing or answering the questions, which in tech pulls top engineers away from more important projects."
"Everyone involved openly jokes—or perhaps secretly fears—that no one reads the responses before tapping the Buy button anyway."
The story here, I think isn't about sales people breathing a sigh of relief. The story I would've pitched would be about how this is the story of humans being lazy and piling a technical "solution" onto something that arguably makes things worse or at the very least doesn't improve outcomes.
Questions now raised:
It is work trying to buy or build something. As a buyer, adding more questions feels like it's reducing risk, but if you don't think properly about those questions, then they introduce inefficiency, noise, and the chance of the best or better suppliers getting through.
This kind of automation is thoughtless, careless delegation that might make things better, that probably helps people feel better during difficult, boring times. But I do think that the work doesn't need to be boring. It certainly needs to be methodical and checked, which some people might feel is boring, in which case, they shouldn't be doing that work!
Ever since these automated text generation engines entered their most recent phase of hype and their rapid adoption into business, I've felt that they are a way to avoid dealing with the underlying, difficult issues of, well, business. They're a way to avoid dealing with bullshit jobs. They're a way to avoid dealing with actually putting time and thought into communication and decisions. And that is understandable, but the point of computing, the point of technology was to free us. We clearly haven't, in most cases, figured out how to successfully, repeatedly, and replicate the use of technology, cheap, networked computing, towards a structure of working people allowing more time for thought and deliberation.
Computing for freedom, not computing for drudgery. And the answer there is that it's always a people problem, not a tech problem. Always.
Shit. I thought this would be shorter. I had the whole plan to see if I could get this one in under 500 words and look! It was not. It is under 2,000 words though!
I maintain it was the RFP thing. The RFP thing alone was ~1,200 words. Ugh. Next week I'm still going to shoot for 500. If I were really serious about this as a writing practice, I would now go and take the time and edit it down to 500. But I am not going to do that today. Sorry.
It's Friday. How's your week been, and how are you?
Things That Caught My Attention is a free newsletter, and if you like it and find it useful, please consider becoming a paid supporter.
Generative AI Is Coming for Sales Execs’ Jobs—and They’re Celebrating | WIRED (archive.is), Paresh Dave, 5 October 2023, WIRED "dot com" ↩
A hackathon! How sweet! ↩
Presumably "Watsonx" because a) adding an x to things is cool (heh, but see the completely foreseeable recent turn of events); and b) Watson turned out to do such a shit job that it did a Comcast and had to change its name. Please don't sue me, IBM, see: What Ever Happened to IBM’s Watson? - The New York Times (archive.is) ↩