How Ray Kurzweil’s 1998 predictions about 2019 are faring

I've visited a dozen countries and I've never encountered an AI system to prevent interpersonal violence. AI gunfire detection system seem to exist solely in a certain country with a tendency to tack startup solutions onto deep-rooted societal issues.

This is a recurring theme throughout the article. The average hospital runs on Windows XP, not a sci-fi network of smart-watches.

Virtual artists are somewhat respected for sure, but AI-generated art is a niche for nerds. I say "art" you think Picasso, not Google Labs.

No, no one passed the actual Turing test yet. The public discussion was doomed from the start, since Turing himself published several variants. This led to an explosion of (mostly dumb IMO) variations that are much easier to solve - which are getting solved.

Honestly, the original predictions are written badly. They are full of bullshit words (many, some, there is) making them not really quantifiable. Also, they contain an implied focus on US tech centers, which allows the existence of a single PR-stunt pilot project as a reason to answer "Totally Right" to anything proposed. Maybe I just don't the point, but this smells like pseudo-intellectual reasoning for "amortality will definitely be achieved before I die, see?"...

> MOSTLY RIGHT

I'll quibble with this one and call it mostly wrong. Yes, a very narrow and specific set of possible cardiac abnormalities can be detected (with significant error rates) by the Apple Watch's optical (PPG) and electric (ECG) heart rate sensors. Other vendors have functionally equivalent capabilities.

But medically speaking, smartwatches are probably decades away from independently diagnosing anything, let along recommending an intervention (i.e. a drug or surgery). Kurzweil's terms have specific meanings in medicine and current technology is nowhere close to that.

What exists today (or, in 2019) is a very narrow use case and far from the broad and general scenario Kurzweil envisioned. And there is no realistic roadmap to achieve any more than incremental development in this area - PPG and ECG were "low-hanging fruit" in some sense, and medically actionable wearable sensors for pretty much any other medical signal present orders of magnitude larger engineering challenges that few people are even seriously working on. Hence, "mostly wrong."

When you instead grade his foresight on the basis of how he predicts the development of technology over time, it's clear that a lot of what would seem to be partial hits if assessed alone should actually be graded as hard fails. Discussion about the ubiquity of computers and their use in non-traditional roles for computers is a firm example here.

> “Many of the life processes encoded in the human genome, which was deciphered more than ten years earlier, are now largely understood, along with the information-processing mechanisms underlying aging and degenerative conditions such as cancer and heart disease.”

This is an example of where the author is going way too easy on Ray (he called it "partly right"). One of the biggest futurist misses of the 21st century is the role of genomics. While genomic sequencing has become incredibly routine, that hasn't actually resulted in the revolution in medicine that it was thought to have by, well, everybody. A key case in point is COVID-19 and some of the mutations: we understand very well details about some of its evolution, and we can pinpoint those mutations to individual genetic letters, but we still struggle to actually develop any predictive effects on its behavior (e.g., is this new strain more infective? more deadly? more resistant against vaccines?--we can't answer those questions without seeing what happens in real infections).

Contrast with organic chemistry. Decades ago, figuring out how to synthesize something ugly like vitamin B12 was insanely difficult, even Nobel Prize-worthy. But nowadays, it's the sort of task you might give to a junior chemical engineer in the first week to onboard them (although maybe not for something as difficult as vitamin B12). The thought was that genomics would lead biology to a similar sort of plane of existence--that we could treat genes as a ready menu for achieving various biological goals, and treating disease (at least those with genetic components) would be handled with a similarly combination of picking and choosing from this well-understood menu. That's what Ray was predicting here (more specifically, that the genetic influence on aging, cancer, and heart disease has succumbed to this process), and we really haven't made any progress on that front.

The ones he was mostly wrong on, are the highly speculative ai boosterism which is why Kurzweil is laughed at by many people.

I've felt grumpy about Kurzweil since 1982 or so. He made some really good innovative stuff happen in computing for blind people and then dived into fame and attention seeking lunacy. He's unquestionably smart, far smarter than me. He appears to have set himself time wasting goals.

I totally fail to see what value he brings google compared to eg Vint Cerf or Rob Pike or Ken Thompson.

This is not exactly true. Welfare in some parts of the US may be minimal; for example, food stamps may be available but not enough to feed yourself every meal. Further, housing assistance may be non-existent or virtually so in the sense that many are homeless while waiting extended periods of time before they can utilize it, don’t qualify because their income is too low (it’s true, but how does that make sense?), or such housing is temporary.

Also, some of the “right” interpretations are rather optimistic or generous in their assessment. But fun read.

The point about facial recognition cameras is fair though.

Thankfully, because the idea that machine intelligence should have rights is based on a fundamental and obvious mistake: first they're not alive, second they've not evolved to feel pain or any other drive related to being alive.

You can't literally kill an artificial intelligence. What if you tried? Joke's on you, I have a snapshot from 10s ago. What if I destroy that snapshot? Joke's still on you, I have a full backup from last night. What if I destroy that backup too? Dude, I got an offlined tape in the vault. And anyway it was distributed to begin with.

Otherwise, you’re in the realms of prophecy or horoscopes where most of the heavy lifting is done by the reader after the fact.

Edited to add my own surprising-but-useless prediction: “Computers as we know them today will be largely abandoned by the year 2050”.

> If “algorithms” is another name for “computer intelligence” in the prediction’s text, then yes...

If that fragment is to be taken seriously, then much of what Kurzweil predicted for 2019 was also true in 1998. Algorithms carried out by computer have been used to administer and surveil since before the 90s.

Personally, my understanding of Kurzweil is that "computer intelligence" is something greater than just an algorithm that is carried out by a machine, but could be done by a human. It would mean something more malleable and conscious than that. Based on that standard I'd say he's wrong about nearly everything (save the privacy one - but again - that was true in 1998 with weaker encryption).

What? How does this follow?

Kurzweil fandom bullshit, all of it.

Life expectancy increases are mostly a statistical fluke of reducing child mortality. Once you factor that out, our upwards trend is way less impressive.

People get into these discussions of specific predictions with their worldviews motivating the evaluation of the predictions. That's why you have strong disagreements about what ostensibly would be a fairly objective measures. Because people see the world in radically different ways.

There are different variations of views towards technology but at the extremes you have "true believers" and what I would call "haters".

What about the cases of virtual divas, such as the OG from Vocaloid, Miku Hatsune, and now the whole section of vtubers, such as Hololive? Seems like it hits that spot, although still a tiny niche.

The narrative of near-future (10-20 years) medical breakthroughs is continually reiterated by the media, only with the horizon pushed further and further forward, with the effect of distorting the public's perception of the actual pace of medical innovation.

I wonder what kind of changes (regulatory or other) you would see if the general public came to accept that radical progress in medicine - absent radical change - wasn't going to happen in their lifetimes, and that medicine twenty years from now would be little different and only marginally more effective than it is today.

> MOSTLY RIGHT

This first one should be “mostly wrong”. The vast majority of decisions are made without any involvement or consultation with machine-based intelligence.

In air combat, whether A2A or air to ground, smart missiles are the primary weapon with planes acting as the launch platform. Guns are there as a backup, to fire warning shots, or to save on the expense of the missile. To wit, according to a former fighter pilot streaming air combat sims, he didn't train much on guns even though it's commonly used in the game for the sportmanship aspect.

Sounds like a lot of song lyrics I've heard.

If I was a pedant, I’d point out that the mRNA vaccines are not antiviral treatments, and this prediction was WRONG; the balance of power clearly lies with potential attackers right now.

I am a pedant.