2023: Is it just me, or was this year especially crazy?

As any of you who’ve already seen my precursor “2024 Look Ahead” piece may remember, we’ve intentionally flipped the ordering of these two end-of-year writeups once again this year. This time, I’ll be looking back over 2023: for historical perspective, here are my prior retrospectives for 2019, 2021 and 2022 (we skipped 2020).

As I did last year, though, I thought I’d start by scoring the topics I wrote about a year ago in forecasting the year to come:

Inconsistently easing semiconductor constraints
Mounting environmental concerns
Unpredictable geopolitical tensions, and
Unclearly legal generative AI

As I also noted last year, maybe I’m just biased but I think I nailed ‘em all. Let me know your opinions in the comments. And here, by the way, are the other topics I ended up not covering in detail in last year’s forecast for 2023 but still briefly mentioned in its summary:

Industry layoffs
Electric vehicles
Autonomous vehicles
The Metaverse, and
Lingering pandemic trends and impacts

In the sections that follow, I’m going to elaborate further on a few of these themes, as well as discuss other topics that didn’t make my year-ago forecast but ended up being particularly impactful (IMHO, of course).

Large language models quickly become commonplace

Generative AI in its various implementation permutations ended up, as I’d predicted a year ago, dramatically accelerating in popularity this year, despite its prodigious environmental-resource impacts (which, as with bitcoin mining’s ongoing transition from “proof of work” to more resource-efficient “proof of stake”, will hopefully decrease over time). Heck, to the power consumption point, some analyst firms are suggesting that generative AI might be what finally kickstarts battery-powered smartphone sales again (as a counterpoint to last-month’s year-ahead forecast of their ongoing atrophy: to be clear, I’m still skeptical).

What I didn’t necessarily predict a year ago though, was the in-parallel emergence and rapid usage increase of large language models (LLMs), which at least currently are unfortunately also environmental catastrophes in their own right. Just a few days ago as I write these words in early December 2023, in fact, OpenAI’s ChatGPT celebrated its one-year birthday. LLMs existed prior to ChatGPT’s unveiling, of course, and other robust implementation options also remain available and under ongoing development. But ChatGPT has captured a disproportionate percentage of the general public’s mindshare, aided in no small part by Microsoft’s sizeable investment coupled with more recent coverage of management turmoil at the company.

In retrospect however, LLMS’ speedy widespread acceptance, both as a generative AI input (and sometimes also output) mechanism and more generally as an AI-and-other interface scheme, isn’t a surprise…their popularity was a matter of when, not if. Natural language interaction is at the longstanding core of how we communicate with each other after all, and would therefore inherently be a preferable way to interact with computers and other systems (which Star Trek futuristically showcased more than a half-century ago). To wit, nearly a decade ago I was already pointing out that I was finding myself increasingly (and predominantly, in fact) talking to computers, phones, tablets, watches and other “smart” widgets in lieu of traditional tapping on screens and keyboards, and the like. That the intelligence that interprets and responds to my verbally uttered questions and comments is now deep learning trained and subsequent inferred versus traditionally algorithmic in nature is, simplistically speaking, just an (extremely effective in its end result, mind you) implementation nuance.

Battery materials as geopolitical chess pieces

In last year’s retrospective, I pointed out the increasing prevalence of batteries based on lithium ion and other chemistries, in two- and four-wheeled vehicles as well as those driven by propellers and impellers, along with other applications. Given that environmental concerns were one of the big-picture topics I’d explored just one month earlier, I of course also noted the importance of ongoing improvements in charge density, cost, charge speed, recharge cycles, and other metrics as a means of meaningfully weaning us off greenhouse gas-generating fossil fuels.

I closed out that section of the writeup with the following comment:

Barring the discovery of vast, cost-effectively mineable new lithium deposits somewhere(s) in the world, we’re going to need to count on two other additional demand-mitigating variables:

Breakthroughs in non-lithium-based battery alternatives, and
Low-cost, high-yield battery raw material recycling

What I admittedly didn’t comprehend at the time was the degree to which the then-current geographic concentrations of lithium and other key raw materials in relatively few regions of the world would find use both, to advance those countries’ leadership in batteries based on those raw materials, and in parallel, to hamper the aspirations of competitors and others. Take, for example, these excerpts from an NPR interview published in late July:

When it comes to supply chains for the electric vehicle industry, China is far ahead for the number of batteries and EV cars that it produces. It’s also cornered the market on the minerals, metal, cathodes and anodes that go into batteries. Can the rest of the world catch up?

The numbers speak for themselves when it comes to critical elements used in electric vehicle batteries and other forms of renewable energy storage. China mines more than two-thirds of the world’s graphite, extracts 60% of the rare earth. It owns almost half of the cobalt mines and controls a quarter of the lithium.

Last year, China refined 95% of manganese, roughly 70% of cobalt and graphite, two-thirds of lithium, and over 60% of nickel. These are all the key materials for lithium-ion batteries that currently dominate the market.

To wit, as announced in late October and effective just a couple of days ago as I write these words, China is restricting exports of graphite. And it’s not just China; a host of African countries rich in various critical minerals are also negotiating tough with the United States and other high-volume importers, for example. The U.S. and others are aggressively seeking out domestic supplies for lithium and other raw materials, but translating a find into high volume extraction won’t happen overnight and may also be constrained by environmental impact concerns.

Political tensions impact technology firms’ businesses

Speaking of China…as of a couple of days ago, the United States issued long-awaited regulations that limit Chinese content in batteries eligible for electric vehicle tax credits beginning in 2024, starting with fully assembled cells and later spreading to raw materials. This is just one example of the suite of technology-related sanctions and other restrictions that the U.S. and other Western countries have issued against China in recent years, seemingly accelerating of late.

Those countries’ officials point, for example, to claimed official China-sanctioned, often even China government-coordinated, espionage programs against Western businesses and political entities, as discussed for example in a recent 60 Minutes segment:

along with price “dumping” designed to force non-China competitors out of markets artificially made unprofitable near-term, only for Chinese businesses to then raise long-term prices once competitors have been eliminated. In bringing up these claims, to be clear, I’m not offering any opinion as to their validity-or-not, I’m just reporting them.

These sanctions, unsurprisingly, also include restrictions on the types, and the performance and other features within a given type, of SoCs and other ICs, along with the optical lithography and other equipment used to fabricate advanced chips. AMD, Intel and NVIDIA, for example, are all export-constrained as to which host processors, GPUs, and AI accelerators, and at what clock speeds, can be shipped to Chinese customers both directly and via intermediaries (NVIDIA in particular appears to have plenty of other customers for its AI-tailored chips and seemingly hasn’t ended up with oversupply). The sanctions have seemingly had at least some effect, at least in the near term, although visionary Chinese firms reportedly stockpiled supplies in advance, anticipating the rules’ unveil. The long-term impact, on the other hand, is less clear.

Ongoing GPU high prices and availability limitations

And speaking of NVIDIA…a year ago, in discussing the forecasted easing of prior pandemic supply chain- and consumer demand-induced semiconductor product constraints, I wrote:

The downturn of the bitcoin mining market has enabled the graphics processor segment (another high-volume consumer of wafers and other fab, test and packaging facilities and resources) to regain some semblance of normalcy, a situation which I suspect will extend into the new year.

I was right…at least sorta…but only for the initial part of the year. Let’s review. Beginning with the emergence of COVID in 2020 and extending into 2022, it was nearly impossible to obtain a board based on a modern graphics processor except at ridiculous markups. Why? Several primary factors:

Pandemic lockdowns, coupled with widespread workers’ illnesses and deaths, crippled supply chains starting from IC fabs all the way to retailer warehouses.
Consumers that had previously inhabited office cubicles during the week instead found themselves sitting on Zoom calls all day. And because they weren’t wasting off-hours time round-trip commuting to the office every day (among other factors), they ended up with spare time on their hands that they filled with (among other things) gaming.
And many of them also delved into speculative bitcoin trading, which was particularly lucrative (at least in places where utility bills were reasonable) if they also did GPU-accelerated bitcoin “mining”.

Unfortunately, the situation as we exit 2023 is eerily reminiscent of recent-past GPU constraints, although the defining factors are different. Folks are increasingly back in the office. And bitcoin trading has fallen out of favor. But AI, as I wrote about earlier, is exploding. GPUs, being massively parallel processing architectures, are well suited for accelerating both deep learning training and inference operations. And AMD and NVIDIA, the leading two GPU suppliers, are both foundry-based from a fabrication standpoint. If you’re them, and you’ve got limited foundry supply at your disposal, what would you prefer to leverage it for: highly profitable AI accelerators or less profitable graphics chips? Exactly.

There’s a specific reason for my showcase of an Intel graphics board in this section, by the way. Here’s the second half of the year-ago paragraph I quoted earlier in the section:

Intel seems to finally be getting its manufacturing house in order, albeit after a multi-year flailing-about delay, which should stabilize (and maximize) yields out of its existing fab network, both for itself and its fledgling foundry services aspirations.

Intel, unlike both AMD and NVIDIA, has less constrained, captive fab capacity available to it. And, although the company’s 2022 foray into bitcoin ICs didn’t pan out, with Intel unceremoniously dumping them a year later (to clarify: the Blockscale ASICs were general-purpose hashing acceleration chips, not bitcoin-specific, although other blockchain-related apps apparently didn’t deliver the demand volumes necessary to rationalize ongoing investment), the company’s re-engagement with discrete graphics has been notably more successful. For the moment, at least, Intel’s products don’t target the high end of the graphics market, but that’s the only segment that AMD and NVIDIA are currently meaningfully active in, anyway. For entry-level and mainstream markets, on the other hand, Intel’s increasingly the only game (pun intended) in town. I’m curious to see how Intel’s pragmatic strategy plays out in 2024 and beyond.

The enduring popularity of HDDs

A couple of weeks ago, Western Digital released two 24 TByte HDD product tiers, with 28 TByte successor versions nipping at their heels. Seagate and other remaining HDD suppliers are making similar capacity-boosting moves. What’s going on? Wouldn’t SSDs’ superior random access performance and lower power consumption (after all, they don’t contain rapidly spinning motors and platters and rapidly oscillating read/write heads), along with steadily decreasing cost/bit metrics, sooner-or-later ensure HDD precursors’ inevitable complete demise?

Maybe that’s what you thought, but I never did, and I’ve got the longstanding documentation to prove it ;-). HDDs have also exhibited steadily decreasing cost/bit metrics over the years. And although they may start out at a higher fixed cost than an SSD, due to the aforementioned motors, platters, read/write arms and heads, and such, beyond a particular aggregate capacity point their total cost ends up being less than that of the SSD alternative (not to mention lower in the total unit volume required to implement that capacity). And regarding power and energy consumption, to quote my favorite engineer-lingo line, “it depends”.

“Cloud” and other enterprise storage is perhaps obviously the dominant driver of HDD demand nowadays, therefore the recently announced WD products’ feature set tailoring. But plenty of consumer NASs (such as the two whirring away downstairs as I type these words) and direct-attached storage devices remain HDD-based, too. And in all these usage scenarios, a higher-performance, lower-capacity flash memory “buffer” may also be included ahead of the rotating media, implementing a “hybrid” architecture. Face it; at the end of the day, we’re all digital (and otherwise) packrats. And HDDs will long have a place in satisfying our accumulated-data needs.

Autonomous vehicle setbacks

Last but definitely not least is the fairly recent story of Cruise’s near-widespread success but rapid demise (near-term, at least) in California, and what it means for the autonomous robotaxi and broader self-driving vehicle market going forward. Let’s review:

In early September, Cruise officials disputed a report that one of their San Francisco robotaxis had blocked the path of an ambulance trying to get a patient (who later died) to the hospital.
In early October, Cruise officials admitted that a pedestrian previously struck by a human-driven San Francisco city bus was then thrown in front of, and run over by, a Cruise robotaxi.
One week later, Cruise competitor Waymo announced that it was expanding its service in that same city.
A couple of days later, Cruise countered with its own expansion news in Houston, TX.
Two weeks after the accident, an interesting news tidbit: the US National Highway Traffic Safety Administration announced that it was opening an investigation into Cruise, for at-the-time unexplained reasons.
Three days later, California’s Department of Motor Vehicles abruptly suspended Cruise’s robotaxi permit, effective immediately, followed by similar action from the California Public Utilities Commission.
And shortly thereafter, we found out why: not only had the aforementioned robotaxi struck the thrown pedestrian, it had subsequently dragged her underneath the vehicle for ~20 feet before finally stopping (whether or not due to remote human operator intervention wasn’t clarified), leaving her in critical condition.

Cruise subsequently lost its robotaxi permit in Los Angeles too, then paused all driverless robotaxi operations to ‘rebuild public trust’. Production of the next-generation Origin robotaxi was abruptly halted. Introduced in January 2023, the steering wheel-less Origin had been claimed “just days away” from receiving the necessary regulatory approval only a few weeks before the San Francisco crash, and Cruise had already assembled a several-hundred-vehicle Origin fleet in anticipation. Pending testing of wheelchair-compatible robotaxis was also halted.

All of Cruise’s vehicles in the field were recalled for software and other updates in early November, and employee layoffs and stock program suspensions predictably followed, along with the resignation of the founder (and with acquiring company GM’s executives taking over). Near-term spending by GM has also been dramatically slashed.

Sounds dire for autonomous vehicles generally, and Cruise specifically, right? Maybe…or maybe not. Waymo, for example, has seemingly come out of its competitor’s troubles comparatively unscathed, at least for now. While Cruise likely flew too fast and close to the sun for its own near-term good, humans’ memories are notoriously short-term. Autonomous vehicles, particularly robotaxis and long-haul trucks, do have compelling benefits and, in these two particular cases (and others), operate in comparatively closed-route and other implementation robustness-beneficial scenarios. And while I’m pretty confident that a human (vs autonomous) driver would quickly stop if he or she sensed another person trapped under a vehicle, people more generally hit other people all the time. I’m not trying to be crass here, just pragmatic; echoing a point I’ve made before, at least some of former Cruise CEO Kyle Vogt’s early-September rant about autonomous vehicles unfairly being held to a different (specifically far more stringent) standard than traditional human-navigated vehicles rings true to me.

Coda

As was the case last year (and plenty of other times before, negatively impacting poor Aalyia’s workload), I’m nearing 3,000 words, with more things that I wanted to write about than I had a reasonable wordcount budget to do so. I’m once again therefore going to restrain myself and wrap up, saving the additional topics (as well as updates on the ones I’ve explored here) for dedicated blog posts to come in the coming year(s). Let me know your thoughts on my top-topic selections, as well as what your list would have looked like, in the comments!

—Brian Dipert is the Editor-in-Chief of the Edge AI and Vision Alliance, and a Senior Analyst at BDTI and Editor-in-Chief of InsideDSP, the company’s online newsletter.

Related Content

2024: A technology forecast for the year ahead
2023: A technology forecast for the year ahead
A tech look back at 2022: We can’t go back (and why would we want to?)
A 2021 technology retrospective: Strange days indeed
<!–

–><!–

div-gpt-ad-inread

–>

The post 2023: Is it just me, or was this year especially crazy? appeared first on EDN.