Lead story - Generative AI - reassessing the risks and use cases
I return from vacation, only to find - more generative AI. But then again, I do welcome the precision of critical analysis. Neil kicks that off with Generative AI in the enterprise - re-assessing the risk factors. Neil defines four fundamental AI tensions that organizations must contend with:
- Deploying models for customer efficiency versus preserving their privacy
- Significantly improving the precision of predictions versus maintaining fairness and non-discrimination
- Pressing the boundaries of personalization versus supporting community and citizenship
- Using automation to make life more convenient versus de-humanizing interactions
In theory, "AI ethics" should help - but doesn't it seem like AI ethics is always lagging behind systems in production? Though Neil has been sharply critical of the problematic field of "AI ethics," he notes some promising developments, including new approaches to operationalizing AI ethics.
Of course, the challenge with generative AI is we can't assess the live enterprise use cases yet. However, George covers one that should go live in 2024: Elsevier wades into generative AI - cautiously. Elsevier has opted not to build its own LLM; it will license ChatGPT instead. But as George writes, this year is all about ensuring good results to research queries:
Elsevier is starting small with an alpha release of the new AI capabilities and taking advantage of its existing citation search engine, knowledge graph, and custom ontology to ground ChatGPT’s results to a chain of trust. This builds on the firm’s previous work on Small Language Models and graph data we covered in March.
Elsevier is also limiting the hallucinatory downsides of ChatGPT by putting a semantic search engine underneath it. He quotes Elsevier:
Using the query that the user types in, we're firing that into a semantic search engine and getting back the list of results. And we're using that, in addition to the query, to prompt the LLM to give essentially a summary. So we're essentially using the LLM as almost the natural language interface.
So when you get the results back, you actually get the references from Scopus that support all of the summary statements that come up in the summary. So that obviously reduces the risk of us making up references because it's very hard to make them up when you've essentially returned them from a search engine.
This strikes me as a well-thought approach to getting the most from an off-the-shelf LLM, while limiting its downsides. But I would also point out that: 1. this organization has considerable data and semantic assets to make this happen, and 2. when an LLM is just part of a well-designed mix, I think you would call this a progression of enterprise tech, not a revolution.
I realized that harshes the buzz of exuberant AI marketing teams everywhere... Just be glad I didn't get a chance to weigh in on George's Why we need to treat AI like a toddler - OWASP lists LLM vulnerabilities (that one came out while I was on vacation; Alex Lee reeled it into their guest edition of Hits and Misses last week).
diginomica picks - my top stories on diginomica this week
- Starbucks digital re-invention - a progress report with a strong AI foundation - Stuart has chronicled Starbuck's digital re-invention since pre-pandemic times. Now another crossroads has arrived: "Experiential convenience powered by digital in an omni-channel way? That’s certainly fully buzzword-compliant at any rate. It’s still early days on this much-vaunted, tech-centric re-invention of Starbucks, but Narashiman’s point about Deep Brew being a foundational platform for change in the current AI-dominant climate is well-made."
- Attention enterprises - generative AI is sexy, but ESG has teeth - Generative AI is all enterprise software vendors want to talk about these days - but are we sleeping on ESG? The regulatory traction behind ESG is going to force the IT spending hand, and a new generation of ESG vendors is on the case. Here's my official ESG about-face. Also see: Madeline's How Panasonic aims to eliminate the equivalent of 1% of total global CO₂ emissions by 2050.
Vendor analysis, diginomica style. Here's my three top choices from our vendor coverage:
- Twilio beats Q2 2023 estimates as CEO outlines ‘Act Two’, focused on AI - Derek: "As most software vendors in the market pursue a generative AI strategy, it will be the ones that utilize their core data the best and build out use cases that are genuinely useful for their customers."
- John Lewis Partnership and Google Cloud - never knowingly under-selling the potential of AI - Stuart on a high stakes retail use case: "This latest expansion will see the struggling UK retailer migrate its tech stack over to Google Cloud and, significantly, tap into the tech firm’s Artificial Intelligence (AI) and Machine Learning (ML) capabilities to help improve the customer experience by enabling the workforce to be more efficient and make better use of data to curate products and services."
- Slack updates its UI to give users more focus time and simplify navigation - Phil's on the case; with Dreamforce around the corner, surely there will be more where this came from soon.
Jon's grab bag - Mark filed an instructional hybrid cloud use case, Hybrid cloud gets the job at UK’s Department for Work and Pensions. Finally, Em filed a terrific green AI piece via AI for plant breeding - the new green revolution?
Best of the enterprise web
My top eight
Zoom Faces Challenges in Navigating the Age of Generative AI – Amalgam Insights's Hyoun Park issued a definitive examination of Zoom's new controversial terms of service. But Park is right; this issue extends far beyond Zoom:
On August 7, 2023, Zoom announced a change to its terms and conditions in response to language discovered in Zoom’s service agreement that gave Zoom nearly unlimited capability to collect data and an unlimited license to use this information going forward for any commercial use. In doing so, Zoom has brought up a variety of intellectual property and AI issues that are important for every software vendor, IT department, and software sourcing group to consider over the next 12-18 months.
In my view, Zoom made numerous mistakes here. As I said to Park on Twitter:
Yes well, to your point, vendors may be able to get away with this kind of tactic on the consumer side where data takeaways are built into the TOS, but I don't believe it's going to fly on the enterprise side, and companies with enterprise ambitions need to do better than this
— Jon Reed (@jonerp) August 13, 2023
Zoom did issue some clarifications on this policy subsequent to Park's post. But as per ZDNet, that may not be enough: Zoom is entangled in an AI privacy mess. Zoom may have stepped it in this time around, but Park is right to extend this issue beyond Zoom:
This is going to be a wild ride over the next year. I’m really looking forward to seeing how @diginomica keeps documenting this shift in software expectations. Between AI shifts & Hashicorp becoming less open source this week, software license agreements are quickly changing.
— Hyoun Park (박현경) (@hyounpark) August 13, 2023
For more of my commentary, click on Park's tweet above. As for Park's note on open source, that brings us to our next pick:
- Why Open Source Matters - with open source at a major crossroads, no one better than RedMonk's Stephen O'Grady to air out the implications: "A world in which non-compete licensing grows at the expense of open source is problematic enough. A world in which vendors blur the definition of open source such that regular users can no longer differentiate between the two is much, much worse."
- Open source controversies - examples of open source at a commercial crossroads are coming fast and furious. Example #1: Oracle, SUSE, and CIQ go after Red Hat with the Open Enterprise Linux Association. Example #2: HashiCorp changes its source licence to BSL.
- Even as cloud growth continues to slow, market remains robust in Q2 at $65B - Ron Miller breaks out the latest cloud numbers and market shares, with AWS retaining its firm lead.
- Should LLMs Write Marketing Copy? - Trick question. If it's copy straight to publication, no. If it's copy to hone and riff on, perhaps. But what sets this New Stack piece apart is the hands-on experiments of the author.
- Misinformation in the Enterprise Technology and Consulting Space - Eric Kimberling's been on a hot streak this year, prying open the topics on services that are too often locked down.
Via Alex Lee, I guess we have to move recipes off the harmless generative AI use case list for a little while:
"the perfect nonalcoholic beverage to quench your thirst and refresh your senses” - chlorine gas recipe nice!
-> welcome to the generative AI "revolution" https://t.co/0lEKPrcA2K
— Jon Reed (@jonerp) August 10, 2023
So I whiffed a bit here:
AI-powered gamification is one of the most abysmal tech "trends" of all time. In a local group, so-called "top contributors" awarded by Facebook are typically amongst the lowest-value, full of viral wiseass + lack of helpfulness.
As an admin I have no control of it - awesome
— Jon Reed (@jonerp) July 27, 2023
Turns out, buried in the bowels of Facebook, I did have some control over these settings for the local group I run. With foot removed from mouth, I stand by the tweet; with a bit more flexibility to create our own badges, we'd actually have something kinda fun. The "top contributor" group badge is brute force - a top contributor, by Facebook's definition, is just a volume award. Whoever posts the most gets the nod. I shouldn't have derived nearly so much satisfaction by turning it off, thereby taking all the top contributor badges from active members away, but I did. Reserve a spot for me in purgatory...
We'll close with my news article title of the week: Florida village terrorized by peacocks plans to use vasectomies to solve the problem. If that seems like an indirect solution, bear in mind they plan to give the vasectomies to the peacocks - at least, I think that's the plan. We'll have to check back in on that one... See you next time.
If you find an #ensw piece that qualifies for hits and misses - in a good or bad way - let me know in the comments as Clive (almost) always does. Most Enterprise hits and misses articles are selected from my curated @jonerpnewsfeed.