Claims

Factual · Field building and careers

“You probably need a thousand dollars to buy a GPU.”

Dario Amodei, OpenAI researcher, interviewed · 21 July 2017 · Dario Amodei on OpenAI and how AI will change the world for good and ill (80,000 Hours)

Getting started in machine learning probably requires about a thousand dollars to buy a GPU.

Mostly true True in substance, wrong in a detail.

The second judge said resolved true; the verdict shown is the first judge's.

What would count as true or false

Judge as of 2017-07-21. The claim is hedged ('probably') and is about a GPU suitable for getting started in machine learning. True: suitable consumer or entry deep-learning GPUs (such as the GTX 1070/1080/1080 Ti or Titan X) or a GPU-equipped setup cost roughly $500–$1,500 in mid-2017. Mostly true: typical cost about $300–$500 or $1,500–$2,000. Mixed: adequate options ranged widely, with many well below $300, or free or cheap cloud GPU access made buying a GPU largely unnecessary, while about $1,000 was still a common figure. Mostly false: a suitable GPU typically cost well under $300 or well over $2,000. False: getting started clearly required no GPU spend, or cost was an order of magnitude different.

Evidence

1. NVIDIA Deep Learning / AI GPU Value Comparison Q2 2017 ServeTheHome · 15 May 2017

   “Right now, if you are budget constrained the NVIDIA GTX 1060 6GB cards are where the value is at.”

2. NVIDIA Announces the GeForce GTX 1080 Ti Graphics Card at $699 TechPowerUp · 28 February 2017

   “NVIDIA today unveiled the GeForce GTX 1080 Ti graphics card, its fastest consumer graphics card based on the "Pascal" GPU architecture, and which is positioned to be more affordable than the flagship TITAN X Pascal, at USD $699, with market availability from the first week of March, 2017.”

3. Pre-configured Amazon AWS deep learning AMI with Python - PyImageSearch PyImageSearch · 20 September 2017

   “For less than $1/hour you can use a machine with a GPU which will dramatically speedup the training of deep neural networks.”

4. Google Colab Google

   “Colab is a hosted Jupyter Notebook service that requires no setup to use and provides free-of-charge access to computing resources, including GPUs and TPUs.”

Reasoning

E2 shows the GTX 1080 Ti, a high-end consumer card suited to deep learning, launched in early 2017 at US$699. That price is inside the criteria's roughly $500–$1,500 range for "true." E1 names the GTX 1060 6GB as the budget-value choice but gives no price, so it does not show that suitable cards commonly cost well under $300. E3, from about two months after the claim date, shows GPU cloud machines were available for under $1/hour, so there was a cheap alternative to buying a card. E4 is undated, so it cannot be placed in mid-2017. The ~$1,000 figure fits the evidence [E2]. E3, from after the claim date, shows that renting a cloud GPU was an alternative, but it does not show that this made buying a GPU largely unnecessary in mid-2017.

Cite this claim

Dario Amodei, “Getting started in machine learning probably requires about a thousand dollars to buy a GPU.” (said 21 July 2017). Verdict: mostly true. Reliability API, claim c_29edbc9c382d, method 0.1.0+c876030c. https://epistacklabs.com/claims/c_29edbc9c382d