Claims

2,653 claims by 33 public figures and organisations, checked against the public record. Each has its quote, source, verdict and evidence. Read them without a key; browse by person.

28,548 claims, sorted by newest said. Choose a claim to read it in full.
VerdictPersonKindSaidClaimTopic
Delivered on time Scott Alexander Promise 22 Jan 2019 Scott Alexander will post his 2019 predictions later in the week of 22 January 2019. Scott Alexander · Promise · 22 Jan 2019 Their own work or organisation
Unresolvable Scott Alexander Factual 22 Jan 2019 At the 95% confidence level, Scott Alexander's 2018 predictions came out only 75% correct. Scott Alexander · Factual · 22 Jan 2019 Their own work or organisation
Not checkable Scott Alexander Not checkable 22 Jan 2019 Please don’t complain that 50% predictions don’t mean anything; I know this is true but there are some things I’m genuinely 50-50 unsure of. Scott Alexander · Not checkable · 22 Jan 2019
Mostly true Scott Alexander Factual 22 Jan 2019 Scott Alexander made predictions in February 2018 for the next five years. Scott Alexander · Factual · 22 Jan 2019 Their own work or organisation
Mostly true Scott Alexander Factual 22 Jan 2019 Cryptocurrency crashed during 2018, after Bitcoin was around $15,000 when Scott Alexander made his 2018 predictions. Scott Alexander · Factual · 22 Jan 2019 Economics and markets
Not checkable Scott Alexander Not checkable 22 Jan 2019 I’m not sure what to think about that. Scott Alexander · Not checkable · 22 Jan 2019
Not researched Scott Alexander Factual 22 Jan 2019 Of his 2018 predictions at 90%, Scott Alexander got 17 right and 1 wrong, a score of 94%. Scott Alexander · Factual · 22 Jan 2019 Their own work or organisation
Not researched Scott Alexander Factual 22 Jan 2019 Scott Alexander makes predictions at the beginning of every year. Scott Alexander · Factual · 22 Jan 2019 Their own work or organisation
Not researched Scott Alexander Factual 22 Jan 2019 Of his 2018 predictions at 70%, Scott Alexander got 14 right and 4 wrong, a score of 78%. Scott Alexander · Factual · 22 Jan 2019 Their own work or organisation
Not researched Scott Alexander Factual 22 Jan 2019 Scott Alexander scores his predictions at the end of every year. Scott Alexander · Factual · 22 Jan 2019 Their own work or organisation
Unresolvable Scott Alexander Factual 22 Jan 2019 Scott Alexander made many correlated cryptocurrency predictions for 2018, which therefore scored badly. Scott Alexander · Factual · 22 Jan 2019 Their own work or organisation
Not checkable Scott Alexander Not checkable 22 Jan 2019 I’m not sure how to deal with those sorts of correlations here except to not make too many correlated predictions. Scott Alexander · Not checkable · 22 Jan 2019
Not checkable Scott Alexander Not checkable 22 Jan 2019 This is partly the fault of not having very many 95% predictions this year, but even so I should have done better than this. Scott Alexander · Not checkable · 22 Jan 2019
Not checkable Gary Marcus Not checkable 21 Jan 2019 This is the most damning paragraph about a major political figure’s loyalties that I have ever read. Gary Marcus · Not checkable · 21 Jan 2019
Not checkable Gary Marcus Not checkable 18 Jan 2019 Handy tool for analyzing utterances from POTUS and his circle. Gary Marcus · Not checkable · 18 Jan 2019
Not checkable Gary Marcus Not checkable 18 Jan 2019 enough already. impeach the liar in chief. Gary Marcus · Not checkable · 18 Jan 2019
Mostly false Gary Marcus Factual 17 Jan 2019 According to Giuliani's disclosure on 2019-01-17, every tweet and statement like the one referenced was a lie, and Trump's campaign colluded. Gary Marcus · Factual · 17 Jan 2019 Other
Mostly false Gary Marcus Factual 17 Jan 2019 A linked report confirms that Donald Trump took talking points from Vladimir Putin. Gary Marcus · Factual · 17 Jan 2019 Other
Not checkable Gary Marcus Not checkable 17 Jan 2019 Democracy in the US is toast if Republicans let this pass. Gary Marcus · Not checkable · 17 Jan 2019
Unresolvable Gary Marcus Prediction 15 Jan 2019 William Barr may become the person who most subverts American democracy, more than Donald Trump. Gary Marcus · Prediction · 15 Jan 2019 AI policy and governance
Not checkable Gary Marcus Not checkable 13 Jan 2019 lying like father and son. and in some ways the single most damning moment in Trump’s presidency. Gary Marcus · Not checkable · 13 Jan 2019
Not checkable Gary Marcus Not checkable 11 Jan 2019 Lindsey Graham, Prophet Gary Marcus · Not checkable · 11 Jan 2019
Resolved true Sam Altman Factual 9 Jan 2019 Sam Altman spoke with Vinod Khosla in an interview or conversation published in January 2019. Sam Altman · Factual · 9 Jan 2019 Their own work or organisation
Unresolvable Gary Marcus Factual 9 Jan 2019 The Kindle edition of The Algebraic Mind: Integrating Connectionism and Cognitive Science is available as of 2019-01-09. Gary Marcus · Factual · 9 Jan 2019 Their own work or organisation
Not checkable Gary Marcus Not checkable 8 Jan 2019 it is a double standard to allow Trump to do so tonight. Gary Marcus · Not checkable · 8 Jan 2019
Not checkable Gary Marcus Not checkable 8 Jan 2019 Throw away your television. Gary Marcus · Not checkable · 8 Jan 2019
Mostly true Sam Altman Prediction 5 Jan 2019 By 2025, AGI will feel within reach to many people in the AI industry. Sam Altman · Prediction · 5 Jan 2019 AI timelines and forecasting
Resolved true Sam Altman Prediction 5 Jan 2019 By 2025, net-gain nuclear fusion will be working at prototype scale. Sam Altman · Prediction · 5 Jan 2019 Technology
Not checkable Sam Altman Not checkable 5 Jan 2019 Predictions for the three most important technological developments that will happen by 2025: Sam Altman · Not checkable · 5 Jan 2019
Unresolvable Sam Altman Factual 5 Jan 2019 As of January 2019, Altman said portfolio 1 in his bet was probably close to satisfying the bet's condition. Sam Altman · Factual · 5 Jan 2019 Their own work or organisation
Resolved true Sam Altman Factual 5 Jan 2019 Sam Altman made a bet almost four years before January 2019. Sam Altman · Factual · 5 Jan 2019 Their own work or organisation
Not researched Sam Altman Factual 5 Jan 2019 As of January 2019, portfolios 2 and 3 in Altman's bet satisfy the bet's condition. Sam Altman · Factual · 5 Jan 2019 Their own work or organisation
Unresolvable Sam Altman Prediction 5 Jan 2019 By 2025, gene editing will have cured at least one major disease. Sam Altman · Prediction · 5 Jan 2019 Technology
Resolved true Jeff Bezos Factual 2 Jan 2019 Mike George, former Alexa VP and long-time Amazon employee, has agreed to lead the Day 1 Academies Fund. Jeff Bezos · Factual · 2 Jan 2019 Their own work or organisation
Resolved true Gary Marcus Factual 2 Jan 2019 Gary Marcus posted "Deep Learning: A Critical Appraisal" on 2018-01-02, asking whether deep learning might be approaching a wall and be just one tool among many needed for AGI. Gary Marcus · Factual · 2 Jan 2019 AI capabilities
Not checkable Gary Marcus Not checkable 2 Jan 2019 hive mind: what progress has been made? what new limits have been identified? what other tools have been invented? Gary Marcus · Not checkable · 2 Jan 2019
Not checkable Jeff Bezos Not checkable 2 Jan 2019 Mike is a true builder who knows how to start from scratch, and he’ll bring tremendous energy, grit, and passion to the job. Jeff Bezos · Not checkable · 2 Jan 2019
Not checkable Jeff Bezos Not checkable 2 Jan 2019 Thank you, Mike! Jeff Bezos · Not checkable · 2 Jan 2019
Not researched Rodney Brooks Factual 1 Jan 2019 As of 1 January 2019, although a few self-driving taxi services had been announced, every one operated with human safety drivers on board. Rodney Brooks · Factual · 1 Jan 2019 Technology
Not researched Rodney Brooks Factual 1 Jan 2019 Brooks said in his 2019 appraisal that he would not edit his original 2018 predictions post. Rodney Brooks · Factual · 1 Jan 2019 Their own work or organisation
Not researched Yann LeCun Factual 2019 Convolution operations can be defined on irregular graphs as diagonal operators in the eigenspace of the graph Laplacian. Yann LeCun · Factual · 2019 AI capabilities
Mostly true Nassim Taleb Factual 2019 In Hayek's account, knowledge is consolidated via prices and arbitrageurs, who trade products, services and financial securities, not binary bets. Nassim Taleb · Factual · 2019 Economics and markets
Not checkable Yann LeCun Not checkable 2019 indicates the number of parameters [22]. Yann LeCun · Not checkable · 2019
Resolved true Yann LeCun Factual 2019 In computer vision, the trend is to apply ConvNets to an entire image to detect, segment and recognize objects of any size at any location. Yann LeCun · Factual · 2019 AI capabilities
Not checkable Nassim Taleb Not checkable 2019 Way too often, the decision science and economics literature uses one as a proxy for another. Nassim Taleb · Not checkable · 2019
Not checkable Nassim Taleb Not checkable 2019 The misinterpretation is still made four decades after Kahneman-Tversky (1979). Nassim Taleb · Not checkable · 2019
Not checkable Nassim Taleb Not checkable 2019 This is typically the case with venture capital: the riskier the project, the less likely it is to succeed but the more rewarding in case… Nassim Taleb · Not checkable · 2019
Not researched Nassim Taleb Citation 2019 Taleb states that Barberis (2003), a review of behavioral economics with emphasis on miscalculation of probability, treats I1 as equal to I2. Nassim Taleb · Citation · 2019 Economics and markets
Resolved true Nassim Taleb Factual 2019 Morgan Stanley correctly predicted the financial crisis but lost $10 billion on its hedge. Nassim Taleb · Factual · 2019 Economics and markets
Not checkable Yann LeCun Not checkable 2019 The important lesson is that Hardware capabilities and software tools both motivate and limit the type of ideas that AI researchers will… Yann LeCun · Not checkable · 2019
Not checkable Nassim Taleb Not checkable 2019 likewise I2, because its construction is similar to VaR, violates the inequality. Nassim Taleb · Not checkable · 2019
Not researched Nassim Taleb Citation 2019 The appendix of reference [9] shows the difference between ensemble probability and time probability. Nassim Taleb · Citation · 2019 Science and research methods
Not checkable Nassim Taleb Not checkable 2019 The well known Kahneman-Tversky result proved robust, but interpretations make erroneous claims from it. Nassim Taleb · Not checkable · 2019
Not checkable Nassim Taleb Not checkable 2019 A binary forecasting record is likely to be a reverse indicator under some classes of distributions or deeper uncertainty. Nassim Taleb · Not checkable · 2019
Unresolvable Yann LeCun Factual 2019 Current model-free RL methods would likely need millions of hours of practice, with numerous accidents, for an autonomous car to learn to drive. Yann LeCun · Factual · 2019 AI capabilities
Unresolvable Yann LeCun Factual 2019 Training a video predictor with a least-square criterion yields blurry frames that average all possible futures. Yann LeCun · Factual · 2019 AI capabilities
Not checkable Nassim Taleb Not checkable 2019 More significantly, the tracking error for probability guesses will not map to that of the performance. Nassim Taleb · Not checkable · 2019
Not checkable Yann LeCun Not checkable 2019 The ability to predict what is going to happen in the world is what allows us to learn to drive without causing accidents Yann LeCun · Not checkable · 2019
Not checkable Yann LeCun Not checkable 2019 Price and power consumption are relatively secondary to performance and flexibility. Yann LeCun · Not checkable · 2019
Not researched Yann LeCun Factual 2019 Modern AI is powered by deep learning, whose origins go back to early experiments with electronic neural nets in the 1950s. Yann LeCun · Factual · 2019 AI capabilities
Not checkable Yann LeCun Not checkable 2019 This model can be used to plan or to train Yann LeCun · Not checkable · 2019
Unresolvable Gary Marcus Prediction 2019 The deep learning approach to language would never work, as argued in a chapter of Rebooting AI. Gary Marcus · Prediction · 2019 Their own work or organisation
Not researched Yann LeCun Factual 2019 Most of today's deep learning systems simply produce an output given an input. Yann LeCun · Factual · 2019 AI capabilities
Not checkable Nassim Taleb Not checkable 2019 Furthermore, one can overestimate the probability yet underestimate the expected payoff. Nassim Taleb · Not checkable · 2019
Not researched Nassim Taleb Citation 2019 Reference [37] supports that for non-binary variables under fat tails, absolute deviations are far more efficient than squared deviations, even when the second moment exists. Nassim Taleb · Citation · 2019 Science and research methods
Not researched Yann LeCun Factual 2019 Python is very slow and memory hungry. Yann LeCun · Factual · 2019 Technology
Not researched Yann LeCun Factual 2019 Facebook performs 3×10^14 predictions per day, some of which are run on mobile devices. Yann LeCun · Factual · 2019 AI compute, hardware and data
Not researched Yann LeCun Factual 2019 Yann LeCun said that after joining in late 1988, he developed the first convolutional neural net and obtained excellent results on handwritten character recognition tasks (zip codes). Yann LeCun · Factual · 2019 AI capabilities
Not checkable Nassim Taleb Not checkable 2019 The methods, for the reasons we have shown here, are highly flawed except in narrow circumstances of purely binary payoffs Nassim Taleb · Not checkable · 2019
Not checkable Nassim Taleb Not checkable 2019 if x moves by up to 1 unit (say, standard deviation), there is a profit, and losses beyond that. Nassim Taleb · Not checkable · 2019
Not researched Yann LeCun Factual 2019 Around 2003, Hinton, Bengio and LeCun, with CIFAR funding, started research projects, workshops and summer schools with the goal of reviving ML community interest in neural networks. Yann LeCun · Factual · 2019 Field building and careers
Not checkable Yann LeCun Not checkable 2019 Obviously, our current learning paradigms are missing a key ingredient. Yann LeCun · Not checkable · 2019
Not checkable Yann LeCun Not checkable 2019 Early neural network chips from Bell Labs. Yann LeCun · Not checkable · 2019
Not checkable Yann LeCun Not checkable 2019 Using FP-32 is necessary because one must be sure that when an experiment fails, it is not because of a lack of numerical accuracy. Yann LeCun · Not checkable · 2019
Not checkable Nassim Taleb Not checkable 2019 So in both cases a forecast of < 1 is correct, but the lumping of the volatility –the fatness of tails– made a huge difference. Nassim Taleb · Not checkable · 2019
Unresolvable Nassim Taleb Factual 2019 The difference between binary and vanilla payoffs is well known in mathematical finance but does not seem to have been passed on to the decision-theory literature. Nassim Taleb · Factual · 2019 Economics and markets
Not checkable Nassim Taleb Not checkable 2019 The Conditional Value at Risk, CVar, measures the expected total losses conditional on default, which is altogether another entity… Nassim Taleb · Not checkable · 2019
Not researched Nassim Taleb Factual 2019 In 2007, after a massive series of foreclosures, the value of most collateral dropped to about one third of its expected value. Nassim Taleb · Factual · 2019 Economics and markets
Not checkable Yann LeCun Not checkable 2019 Many problems are difficult to represent with fixed-size tensors or variable-length sequences of tensors, but are better represented by… Yann LeCun · Not checkable · 2019
Not checkable Nassim Taleb Not checkable 2019 which can allow in absurd cases p∗ to exceed 1 when the distribution is grossly misspecified. Nassim Taleb · Not checkable · 2019
Not researched Nassim Taleb Citation 2019 Cited sources [13], [30], [31] show that variability of variance (heteroskedasticity) produces higher kurtosis in the resulting distribution. Nassim Taleb · Citation · 2019 Science and research methods
Not checkable Nassim Taleb Not checkable 2019 The difference is exaggerated when the variable under consideration lies in the power law classes of probability distributions. Nassim Taleb · Not checkable · 2019
Not checkable Nassim Taleb Not checkable 2019 An economic "collapse" or a climate "disaster" must not be expressed as an event in {0, 1} when in the real world it can take many values. Nassim Taleb · Not checkable · 2019
Not researched Yann LeCun Factual 2019 DL architecture research is evolving toward more sophisticated and dynamic network architectures, and inputs and internal states that are graphs annotated with numerical objects. Yann LeCun · Factual · 2019 AI capabilities
Not checkable Yann LeCun Not checkable 2019 In effect, dynamic DL systems can be seen as the execution trace of a program, with conditionals and loops that are input-dependent. Yann LeCun · Not checkable · 2019
Not checkable Yann LeCun Not checkable 2019 Multi-Scale DenseNet with conditional computation Yann LeCun · Not checkable · 2019
Not checkable Nassim Taleb Not checkable 2019 as games are necessarily truncating a dimension from reality Nassim Taleb · Not checkable · 2019
Not checkable Nassim Taleb Not checkable 2019 Probability is just a kernel inside an integral or a summation, not a real thing on its own. Nassim Taleb · Not checkable · 2019
Not researched Nassim Taleb Factual 2019 Financial regulators use erroneous approximations of CVaR, a risk-management flaw that may have caused the 2008 crisis. Nassim Taleb · Factual · 2019 Economics and markets
Not checkable Nassim Taleb Not checkable 2019 In fact under fat tails, there is no "typical" collapse or disaster, owing to the absence of characteristic scale, hence verbal binary… Nassim Taleb · Not checkable · 2019
Not checkable Nassim Taleb Not checkable 2019 Succession 1 (thin tails): {1, 1, 1, 1, 1, 0, 0}. Mean variation= 0.71. P/L: g(x) = 2. Prediction is correct. Nassim Taleb · Not checkable · 2019
Not checkable Nassim Taleb Not checkable 2019 We can finally assess the speed of convergence, the rate at which higher moments map to those of a Gaussian distribution Nassim Taleb · Not checkable · 2019
Not checkable Nassim Taleb Not checkable 2019 Decision making is sequential. Nassim Taleb · Not checkable · 2019
Not checkable Nassim Taleb Not checkable 2019 Note that we are operating under the mild case of known probability distributions, as it gets worse under parametric uncertainty. Nassim Taleb · Not checkable · 2019
Not researched Yann LeCun Factual 2019 On average, neurons in the brain are at 2% of their maximum activation. Yann LeCun · Factual · 2019 Science and research methods
Not checkable Nassim Taleb Not checkable 2019 almost always have their valuations react in opposite ways when one increases the kurtosis of Nassim Taleb · Not checkable · 2019
Not researched Yann LeCun Factual 2019 In the early days, training a simple ConvNet for handwritten character recognition could take weeks. Yann LeCun · Factual · 2019 AI compute, hardware and data
Not researched Nassim Taleb Factual 2019 As volatility rises, a binary option struck at X0 goes to zero in value while a standard call of the same strike rises toward the asset's level regardless of strike. Nassim Taleb · Factual · 2019 Economics and markets
Not checkable Nassim Taleb Not checkable 2019 arguments for the consolidation of beliefs via prices does not lead to prediction markets as discussed in such pieces as [25] Nassim Taleb · Not checkable · 2019
Not checkable Nassim Taleb Not checkable 2019 One suggestion by the author is for the M5 competition to correct for that by making "predictors" predict the minimum (or maximum) in a… Nassim Taleb · Not checkable · 2019
Back to the list

Factual Their own work or organisation

“Since I made a lot of predictions about cryptocurrency and all of them were correlated, this went badly.”

Scott Alexander, blogger 22 January 2019 2018 Predictions: Calibration Results (Slate Star Codex)

Scott Alexander made many correlated cryptocurrency predictions for 2018, which therefore scored badly.

Unresolvable The public record could not settle it.

What would count as true or false

Judged as of 2019-01-22. True: Scott's 2018 prediction list included several cryptocurrency predictions that depended on the same underlying trend (for example, prices staying high), and most of them resolved wrong in his scoring post. Mostly true: there were several such predictions, but they did somewhat better than 'badly' implies, or 'many' means only a few. Mixed: about half of the crypto predictions failed. Mostly false: there were few crypto predictions, or most of them came true. False: there were no crypto predictions, or all of them came true.

Evidence

  1. Predictions For 2018 Slate Star Codex (Scott Alexander) · 6 February 2018 Own record
    “17. Bitcoin is higher than $5,000 at end of year: 95%”
  2. Predictions For 2018 Slate Star Codex (Scott Alexander) · 6 February 2018 Own record
    “18. Bitcoin is higher than $10,000 at end of year: 80%”
  3. Predictions For 2018 Slate Star Codex (Scott Alexander) · 6 February 2018 Own record
    “Bitcoin prices are already too high (they were higher when I wrote these predictions a few days ago).”
  4. 2018 Crypto Year in Review | CoinCodex CoinCodex · 3 January 2019
    “BTC would eventually reach as low as $3,200 on December 15.”

Reasoning

Scott Alexander's 2018 list includes at least two linked Bitcoin predictions: 95% that Bitcoin ends the year above $5,000 (E1) and 80% that it ends above $10,000 (E2). An independent source reports Bitcoin fell as low as $3,200 on December 15, 2018 (E4). That source does not give the year-end price, and the scoring post is not in the evidence. The evidence therefore does not settle whether most crypto predictions resolved wrong, or whether there were 'many' of them.

Cite this claim

Scott Alexander, “Scott Alexander made many correlated cryptocurrency predictions for 2018, which therefore scored badly.” (said 22 January 2019). Verdict: unresolvable. Reliability API, claim c_dd27ea9d38fa, method 0.1.0+8bea457b. https://epistacklabs.com/claims/c_dd27ea9d38fa

Verdicts are made automatically; how verdicts and scores are made. The same data is on Hugging Face under CC BY 4.0.