Layoffs & LLMs

Thanks everyone for following the Win, Lose, or Draw newsletter and for all the inspiration. Most importantly, thanks for all the content notes. This week these two topics caught my eye and seem well worth the effort given the current trends:

1. Why are there so many layoffs? What's behind this trend?
2. What are LLMs and how will it grow the industry?

The backlog is growing and it's definitely been fun to hear from folks looking for this kind of research insights. For those who are looking for insights after the last couple newsletters, I've added the Following up on Last week and Other Resources sections towards the end of this newsletter. Thanks everyone!

Layoffs

First, on this topic, let me just say my heart breaks for families who have had someone lose their job to a layoff. Layoffs suck! I hope that folks find new work quickly.

While some companies have been accused of labor hoarding during the pandemic, 2023 has been inked as the year of layoffs where many organizations pushed for greater efficiency due to the economic climate. Adding to this trend, Gen AI's incredible adoption and the need to create new hiring plans by making space for AI skills. My curiosity has me wonder why organizations opt for layoffs and how do they help? Let's learn more:

Win

• According to Why are there so many tech layoffs, and why should we be worried? by Stanford University Graduate School of Business , layoffs are just a bad idea and seem to happen at the same time across many organizations due to contagious behaviors. Agreed. Recent layoffs by many of the tech companies suggest that their hiring practices as well as their layoffs may exhibit this exact correlation to contagious behavior, with data from Layoffs.fyi showing an interesting correlation in January 2023 when 谷歌 , 亚马逊 , 微软 , IBM , and Salesforce all had significant layoffs within days of each other. Also mentioned in the Stanford article, companies that avoid layoffs may gain competitive leverage. Hey 苹果 - I'm looking at you with your controlled hiring and lack of significant layoffs for the last couple years!
• In Generative AI and the Future of Work in America by 麦肯锡 , while layoffs have been making headlines, there is significant demand for better automation and advanced skills which will lead to greater hiring in the coming years. Agreed. Forward momentum requires concepts like creating and using AI capabilities which require significantly skilled labor. For this reason, it is worthwhile to upskill take advantage of all the opportunities around you.
• Fortune 's article Landing a new job will take you 3 months in this economy has several interesting insights from ZipRecruiter which seem to match other industry data. In this article, the labor market is still consider "tight" and a majority of job seekers tend to land within 3 months. Agreed. Based on this and other data, an active job seeker will find a new job but companies are being more selective in some industries. If you want to avoid companies that engage in layoffs, consider evaluating Revenue per Employee as a key metric with trends as a good prediction for how the organization manages its talent.
• In Deloitte Is Using AI to Reduce Layoffs. You can too. by Inc. Magazine , Deloitte is said to be planning for greater work innovation using AI to avoid layoffs. Understanding value creation and determining where to upskill staff before the organization encounters challenging times may prove useful. Agreed. In some ways, layoffs point at poor management which could be better supported through systems. Instead of moving towards a layoff to open positions for AI, maybe some of the staff loyal to the company would dearly love a new opportunity to brighten their futures.
• According to Cyber Skills Gap Reaches 4 Million, Layoffs Hit Security Teams by Infosecurity Magazine , a significant amount of layoffs have hit the once bulletproof cybersecurity industry. Agreed. Lots of folks are popping up with layoff concerns as well. The primary driver for most of these layoffs is economic uncertainty and the quest for a change to Zero Trust architecture. Vendors are moving towards capability consolidation to bring all the related security features together and reduce the integration burden on customers. Similarly, some security jobs are moving into the business with security skills needing to be re-aligned with product development.

Lose

• In the article, Companies are using layoffs to cut new-hire salaries and win back the power they held before the pandemic, some recruiters say by Business Insider , some companies are taking longer to hire and using layoffs to reset salary expectations. According to the article, some organizations are taking a "take it or leave it" attitude towards hiring, re-establishing their power when it comes to negotiating work and pushing back on talent. Disagree. While it seems intuitive, there is very little data that supports this view. Most organizations have expressed difficulty in finding talent despite the surge in layoffs. More importantly, it is hard to agree without some sort of data to back up these claims.

Draw

• In SAP is restructuring 8,000 jobs as it shifts focus to AI by CNN , SAP is restructuring 7% of its workforce and states that it is not a layoff but that staff will have the option to be redeployed and trained or offered voluntary redundancy. Draw. With the report mentioning that SAP is intended to have roughly the same headcount by the end of the year, it means that space is likely being made for new AI skills to enter SAP as a must have. This comes at a time when AI resources are challenging to hire so it might mean that by the end of the year they actually might have fewer headcount and require greater efficiencies. Waiting a year to see how this story plays out isn't too long to wait. It's clear though that while cloud transformation was a major disruptor, AI might change the game.

If you've experienced a layoff recently and looking for a little advice, this article from The Muse on 5 People You Should Reach Out to After You’ve Been Laid Off has a few nuggets to consider. For others, definitely feel free to share any resources in the comments to help folks that have been through a recent layoff to get back to work.

LLMs (Large Language Models)

Ok, let's jump over to LLMs. A quick primer on What Are Generative AI, Large Language Models, and Foundation Models? by Georgetown University .

I know AI is all the rage these days. (pun intended) Probably the big question on everyone's mind: How will LLMs become profitable at scale? Sure, new AI capabilities are popping up like wildfire with many creating revenue streams from generated output, but how will creators actually benefit in the long run? Will creators stop using some platforms because they are being used to mine their creative works? Or maybe hackers will step in and discover methods that could be used by creators to protect themselves? Will Big Regulation have to step in where Big Tech won't?

I guess we'll see where it settles once the bulk of the lawsuits get adjudicated. On that note before I get into the Win, Lose or Draw analysis on LLMs, I joined up with Mark Miller and Joel MacMull to launch AI, the Law, & You . Check out this week's episode: AI Copyright Law for Non-Humans . I'm definitely excited about the series!

Win

• In Are LLMs Content Thieves? by Bob Roitblat , he raises a provocative point that LLMs ingest data to inform their training scraping up what they can to increase their value but do not go so far as to display content. Agreed. More importantly, he raises the point that we must all get much more informed about what LLMs because understanding how LLMs operate is essential for a well-formed debate. That said, the title of the article is what drew me to it and I think he is right that we need to determine if theft has taken place or not. And, this is where I think the biggest issue exists, in that once again, security and safety will be required to make LLMs fully valuable. For this reason, anyone working on Content Security and Licensing will likely see a very bright future ahead, because securing what's valuable makes it worth paying for.
• According to How to Hack Large Language Models by Infosecurity Europe , there are already many types of weaknesses for LLMs, such as: Prompt Injection, Prompt Leaking, Data Training Poisoning, Jailbreaking, Model Inversion Attack, Data Extraction Attack, Model Stealing, and Member Inference. Agreed. In reality, I think there will be many more attack vectors that surface as this set of capabilities permeates its way into technology. The challenge with LLMs is that data governance in most organizations has yet to mature enough so that its properly handled like the jewel that it is. At its root, the value in an LLM is based on the data and relationships around it. An LLM without good input isn't worth much. For this reason, the current behaviors of AI platforms suggest that as much as it is disruptive, it will itself be disrupted by the desire to fully monetize the value in artificial intelligence.
• Maybe John Willinsky was ahead of his time with his book Copyright's Broken Promise: How to Restore the Law's Ability to Promote the Progress of Science . In this book, John explains how copyright law is becoming antiquated by the progress of science and technology and potentially even holding humanity back from saving itself. Agreed. The challenge with LLMs and other AI is commonly steeped in the evolutionary entanglement between humans and machines. Can we collectively think simply about what is needed when we don't and may never all understand the complexity of scientific breakthroughs? To sum up the discord, humans need the ability to be paid to live and computers require cheap input to solve broadly. In the end, a network effect will eventually be discovered and it will likely correct the market.

Lose

• In The ‘Don’t Look Up’ Thinking That Could Doom Us With AI by TIME , the article discusses the need to evaluate what is being said about AI potentially causing the end of humanity and how simply dismissing the concern could be dangerous. While the article raises lots of good points, it states that there are obviously much smarter AI architectures making reference to Einstein vs. GPT4. Disagree. Instead, what I think is being missed by most is not that LLMs are being used to compete with a single human but instead taking all of what humans know and using that to become smarter than all of us combined. A meme popped up today that said: "Only way to keep AI from taking your job, give it memes and hoard your secrets." What concerns me more is alluded to in the article but is more pointed in AI creates new environment injustices, but there's a fix by University of California, Riverside .

Draw

• According to 5 Reasons People Get Laid Off by Harvard Business Review , LLMs is on their list as the #1 reason why some tech workers are currently being displaced in a layoff. Draw. Sure, being relevant and keeping up with the demand for skills is important but it's not clear whether LLMs are so hot that it is worth displacing workers who could learn. LLMs still need to get a lot better before this type of rationale makes sense but it is certainly possible that the perception is that companies can do better. I guess my work spent helping companies migrate to the cloud has me quite a bit skepitcal.

Following up on Last Week

A quick recap on last week's Win, Lose, or Draw on Remote Work and Cybersecurity Regulations , some folks wanted insight and deeper correlation on how Remote Work and Cybersecurity Regulations might relate to each other:

• With a significant talent shortage in cybersecurity still looming pressured by growing demand from regulatory pressures, remote work may be the best option for hiring cyber talent moving forward.
• Evaluating cybersecurity tasks and jobs to determine whether work is required to be supported in-office is a good way of balancing value creation.
• Cybersecurity talent should seek to upskill and continue to learn about new capabilities when possible.

And based on this week's Win, Lose, or Draw on layoffs, perhaps we're going to see companies spend even more on attracting cyber talent. Sadly, it only means that these organizations are gambling that adversaries won't notice a reduced investment in cyber.

Other Resources

Want to collaborate?

Feel free to suggest a topic, collaborate, or let me know about your content for an upcoming edition by sending a DM to Thirdscore. Follow our page to find out more about Thirdscore projects and our upcoming launch plans.