Mondays Pick Up Speed Faster Than We Expect
Thank you all for subscribing. This week I have gotten a lot of quality feedback in emails and DMs. Your time and the conversations you start are incredibly valuable. I also appreciate the likes, which help me select topics to keep developing content around.
I have restarted the third-party content share posts, with the first coming out last week. I will try to keep them coming if there is content worth sharing. My goal is to surface the gems that are not well shared.
This post, a semi-technical explanation of social media recommender models, is a good example. There are thousands of posts addressing this exact topic. What makes this one unique is the target audience, regulators, and politicians. It’s a peak behind the curtains that shows the types of content think tanks create to influence policy decisions.
This whitepaper looks a lot like research, but it’s a position paper with similar aims to the last piece. While it seems to target social media companies, the true audience is policymakers. The author provides content moderation options that have long been adopted by social media companies but are poorly understood by regulators.
One of their sponsors? Meta. They employ their own lobbyists and think tanks to shape regulators’ perceptions. This back and forth is rarely covered in depth, so these articles are essential.
The UK government posted an update to their analytics dashboard with a hidden gem. The dashboard benefits section listed the reasons for making their updates. It’s an excellent case study of what many data organizations can do to make their data products more usable.
In all great data products, iterations and improvements focus on removing steps. That’s what the UK’s updates highlighted. The fewer steps, the higher the adoption rate goes. Data product design and user experience use cases are hard to find.
Recession risks are rising, and many are dusting off their lessons learned from the Great Recession. This post revisits a 2010 study. Companies that invested in the “right” innovation initiatives came out of that recession with higher profitability and performance. They looked at companies emerging from the pandemic and found similar trends.
I have spent the last month in budget and strategy planning meetings with clients. I see a trend they also noted. Uncertainty and the potential economic downturn drive demand for data and AI. Improving productivity and decision support are the umbrella use case categories I see the most interest in.
Companies plan to hire more data professionals, increase investment in automation (buying and building), and build holistic data and AI strategies. My clients are biased towards my recommendations, but I also hear the same sentiment from other data organizational leaders in my network. Articles like this one provide further evidence that the trend is real.
There is massive interest in platform and vendor consolidation. That’s the process of replacing multiple systems with a single company’s offerings. Oracle, SAP, Crowdstrike, and Microsoft are benefiting from the move to simplification. Numerous systems that don’t interoperate well are a primary barrier to maturity, so consolidation will help businesses mature faster.
Data and AI strategies are C-level concerns. Their attention drives action throughout the business to align initiatives and transformation to drive value creation. Data and AI maturity will benefit from this trend as well.
Companies are looking for ways to move from the prototype phases to scaling and handling more use cases. Everyone wants speed and efficiency, which comes with a higher price tag. Even in the downturn, many companies are willing to invest.
In 2008, technology was the victim of the earliest and deepest cuts. I believe that was the result of businesses being blindsided by the Great Recession. Few saw a deep downturn coming, and few were prepared.
Even with the high levels of uncertainty and volatility, businesses had 6+ months to prepare this time. No one is positive about which specific scenario will play out, but all companies are adapting their strategy and spending.
At the same time, data science is maturing. Models have higher business utility than ever. The tools and infrastructure are available to support more practical applications. Large-scale models and causal methods are introducing advanced capabilities.
GPT-3 is replacing low-end copywriters and will disrupt content creation even further in the next year. Content companies haven’t internalized all the practical applications yet. When they adopt at higher rates and operationalize more complex use cases, that’s when the implications will be more evident. I see many low-quality content publishers and creators going away altogether.
Microsoft’s aim with Copilot is to discover and resolve defects autonomously. That will be game-changing because software developers will be much more productive. Copilot could take away the lowest value-generating activities and allow developers to focus nearly all their time on building. ML and Data Engineers will benefit too.
The most significant value proposition for Copilot is combining it with diagnostic models. Discovering failures and recommending fixes are high-value automation. Microsoft’s sitting on a goldmine.
Stable Diffusion is disrupting the stock photos and illustration industries. I can get an image of almost anything in seconds at an extremely low cost. The newest wave of vision models is also being updated to handle animation. Those capabilities are not far off.
Some advertising teams are connecting image generation APIs with their ad recommenders. They are running early pilots to autogenerate highly personalized ad images in real-time. When simple, short autogenerated video becomes possible as well, that’s another disruptive use case.
Consolidation vendors are advancing high-value AI applications. They benefit from the business trend and put AI-supported tools into more businesses. AI is more accessible to people on the front lines.
The workforce is slowly becoming model literate, and those at the leading edge of the learning curve are identifying high-value opportunities. Jeff Bezos said the people closest to the problem are the best ones to solve them. Putting advanced technology within reach and training people to understand it will accelerate solutions discovery.
The pandemic pulled digital adoption forward by 5 years, and this downturn could have similar impacts on AI. Digital use cases were confined by the logical nature of software solutions. AI doesn’t have the same bounds.
The critical paradigm shift from software to models is flexibility. Traditional software doesn’t hop use cases like GPT-3 or Stable Diffusion can. Anyone with a smart home assistant has learned to ask whatever comes to mind and see if it works. A model literate workforce will soon implement the same thought process.
Software use cases progress linearly. AI is exponential. One new model can have hundreds of applications. With minimal retraining, that could expand to thousands. How fast will OpenAI get to a massive valuation? I see a path to them becoming one of the most valuable companies in the world. Competition and the rate of progress will be their biggest challenges.
The demand for data science is going to follow a similar growth rate. Consolidation and automation will alleviate some of that, but not enough. Data organizations will be forced to prioritize relentlessly to make up for the shortage of people. Optimizing opportunity selection and shortening development cycles are critical.
AI is at the end of the beginning. The technology is ready to take massive steps into the real world and successfully integrate into thousands of practical applications. The pace will catch most businesses off guard.
This isn’t the fastest business and technology has ever advanced. Progress and advancement are the slowest we will see for the rest of our lives. Today, companies are less reliant on technology than they will ever be again.
AI is getting ready to snap its fingers, and businesses we see as invincible today will be gone by 2025. The companies that will be born and grow during the same timespan will be amazing. Zombie companies need to die off so progress can be made without their drag. Downturns and conflict accelerate innovation and progress. The companies we lose will be overshadowed by what we gain.
And you are all at the leading edge of opportunity. Enjoy the ride.