ICML 2018 Registrations Sell Out Before Submission Deadline

In a shocking tweet, organizers of the 35th International Conference on Machine Learning (ICML 2018) announced today, through an official Twitter account, that this year’s conference has sold out. The announcement came as a surprise owing to the  timing.  Slated to occur in July, 2018, the conference has historically been attended by professors and graduate student authors, who attend primarily to present their research to audience of peers. With the submission deadline set for February 9th and registrations already closed, it remains unclear if and how authors of accepted papers might attend.

One lucky ICML hopeful, Regus Terring, a recruiter at Goldman Sachs, took to Twitter to celebrate his successful registration. Tweeting under the handle @disruptivesynergy, he claimed that the available tickets to ICML cover only 12% of demand at that price. If true, then to our knowledge it’s the first time in the history of computer science when registration at a top-flight conference proved more selective than publishing in its proceedings.

While the final accounting hasn’t come in yet, a preliminary analysis based on social media postings suggests that nearly all available tickets were claimed by recruiters, science managers, and startup “CEOs”, all looking for elusive machine learning unicorn engineer/scientist.

Blandé Bloviateur, the founder of NeuronFab, a Paris-based 3D-printing startup developing a platform for fabricating self-replicating rectified linear units, put it thusly:

ICML and NIPS, you find the unicorns that can GPU deep learning and convex the non-convex optimisation with tensors are big and you engineer distributed clouds in the blockchain because it’s just pure mathematics and also deep learning is maths but growth starts with research for deep models and big data opportunities! we hire edge TPU and also 3PU EMPs, as well as other verticals.

When asked who he might recruit if no engineers managed to register, Bloviateur offered me a t-shirt and a branded Rubik’s cube.

Reached for comment, machine learning luminary and ICML 2018 General Chair Francis Bach commented that something needs to change. However, he cautioned that he would be unable to express this sentiment in the post-conference debrief because he too was unable to secure a ticket this year.

The overwhelming interest in ICML follows a trend that has been building over the last few years. Following deep learning breakthroughs in 2012, popular interest in deep learning erupted. With interest reaching far beyond the ivory tower of academia, attendance at academic machine learning conference has taken off as well. Last year, the Neural Information Processing Systems conference (more commonly called NIPS), sold out in October, with over 3200 participants  registering more than one month before the December conference.  Following the pattern, 2017 NIPS registrations proved even more electric with the main conference selling out in September.

As the pattern continues, what researchers will do with their travel budget, and what will transpire in a conference without researchers remain open questions.

Author: Zachary C. Lipton

Zachary Chase Lipton is an assistant professor at Carnegie Mellon University. He is interested in both core machine learning methodology and applications to healthcare and dialogue systems. He is also a visiting scientist at Amazon AI, and has worked with Amazon Core Machine Learning, Microsoft Research Redmond, & Microsoft Research Bangalore.

5 thoughts on “ICML 2018 Registrations Sell Out Before Submission Deadline”

  1. Very easy solution, if they say the price doesn’t match the demand, increase the price for non-academics. Simple objective function adjustment 🙂

    1. Increasing the price – as someone suggested – is optimizing for the wrong thing.

      You introduce something the equivalent of business class airfare. Certain companies will be happy to pay for that for certain employees without thinking twice about price, and it depends not just on conscious decisions but also on how easy it is to hide these costs when getting reimbursed. Other companies (like startups) would and should pay for only economy travel. But flying business class vs economy is about the convenience of your employee, not about how useful it is for them to attend a conference, or how useful it is for the community that they are there.

      By increasing the price you can indeed control the market but you also artificially engineer the audience. “this company is happy to pay arbitrary amounts of money to send you to NIPS” could be a slogan you would then put on a job advert.

      Also, what would be a reason for a company to shell out a lot of money to send someone to NIPS or ICML? How does someone working in industry justify their attendance: the easiest thing companies understand is hiring: it’s easiest to measure the ROI on NIPS as the the number of job candidates you have met or added to your database. By this logic, it makes more sense to send someone good at recruiting over someone who actually understands the content let alone gives you some interesting comments or things to think about when you present your poster.

  2. ROFLed, thank you! so true ;-)))
    “3D-printing startup developing a platform for fabricating self-replicating rectified linear units” – I have to get credit and invest into this startup asap.

  3. This experiment in fake news has now appeared as “Data Science News” in the NYU Data Science newsletter I receive weekly. It’s quite strange how educated people are taking this satirical article at face value after blaming all those “deplorables” for believing fake news in the recent US presidential election. Seems like we all have our versions of fake news which we are gullible enough to believe without simple fact-checking.

Leave a Reply

Your email address will not be published. Required fields are marked *