No, we did not just upload memories matrix style.

posted in: Neuroscience, Science | 0

Catchy headlines have been whizzing around the internet recently:

Scientists develop Matrix-style technology capable of ‘uploading knowledge’ to your brain” – The Independent

Scientists discover how to ‘upload knowledge to your brain’ ” – The Telegraph

The news reports claim trained pilot’s had their brain activity monitored and that activity was then “transmitted”/uploaded to naive participants learning to fly in a simulator. But I am sorry to report none of this Matrix style memory and skill “uploading” is true.

Although it is something we all hope for, and one day may be true, the tech isn’t here yet. Interestingly, the actual study produced, published in a prominent scientific journal Frontiers in Human Neuroscience, is a pretty good study (outlined below) and does not make any outrageous or overblown claims.

However, this study comes out of a private commercial laboratory, HRL Laboratories LLC, owned and operated by The Boeing Company and General Motors, who issued a press release (below) along with the journal publication. It seems to be at this press release level that everything breaks down.

MALIBU, Calif. February 10, 2016 — It’s a case of life imitating art. Much as the sci-fi film “The Matrix” depicted a device capable of enhancing skill acquisition, researchers at HRL Laboratories, LLC, have discovered that low-current electrical brain stimulation can modulate the learning of complex real-world skills.

Dr. Matthew Phillips and his team of investigators from HRL’s Information & System Sciences Laboratory used transcranial direct current stimulation (tDCS) in order to improve learning and skill retention. “We measured the brain activity patterns of six commercial and military pilots, and then transmitted these patterns into novice subjects as they learned to pilot an airplane in a realistic flight simulator,” he says.
The study, published in the February 2016 issue of the journal Frontiers in Human Neuroscience, found that subjects who received brain stimulation via electrode-embedded head caps improved their piloting abilities. “We measured the average g-force of the plane during the simulated landing and compared it to control subjects who received a mock brain stimulation,” says Phillips.

While previous research has demonstrated that tDCS can both help patients more quickly recover from a stroke and boost a healthy person’s creativity, HRL’s study is one of the first to show that tDCS is effective in accelerating practical learning. Phillips speculates that the potential to increase learning with brain stimulation may make this form of accelerated learning commonplace. “As we discover more about optimizing, personalizing, and adapting brain stimulation protocols, we’ll likely see these technologies become routine in training and classroom environments,” he says. “It’s possible that brain stimulation could be implemented for classes like drivers’ training, SAT prep, and language learning.”

Perhaps the company was trying to build hype and make its study viral by lying about the actual results in their press release. Further breakdown happened journalistically, where “science writers” took the press release on face value and did not bother even reading the study’s abstract where they would have found very little evidence for the claims made in the press release.

Update March 10, 2016: HRL Laboratories media team has replied to my emails (full correspondence at the end of this post). They have cleared up a few things – mainly, commercial and military pilots did indeed have their brains monitored but this data is being used in a separate manuscript(s) and was not used in the Frontiers of Human Neuroscience article. They have since added a small line to the bottom of the press release which reads,

“The results from the expert brain patterns are the subject of subsequent manuscript(s).”

A small victory I suppose. However, many questions still remain as to what stimulation protocol was used for naive participants and on what data they make the claim that practical learning was accelerated. It is still perplexing why the lead author claims that these expert pilot patterns were “transmitted” to novice subjects when the data from expert pilots has nothing to do with the article in question.

Frontiers has informed me via twitter they “were not involved in HRL’s press release” and commented on HRL’s fraudulent press release that, “Yes it [is] very unfortunate. All research is exciting enough as it is, “sexing-up” is never necessary.”

 

So what did the study actually find?

First a critique:

The scientific article about the study shows that it is actually a pretty good study. The only critique I have is that they used employees of HRL Laboratories as test subjects and had quite a small sample size as a result. This causes two problems. 1) Since they are employees, the subjects will want the study to work and may be extra focused and motivated. This is different from the usual relatively unbiased university student participating in studies for money or class credit. The employees of the company won’t want to disappoint their colleagues (this is called Demand Bias) and may respond differently to the tasks from how they would naturally (this is called Response Bias).

The scientific article authors are not naive to this and both point it out in their article and take recommended measures to mitigate these biases. Good on them. Although, as we will find out below, the quote in the press release by the senior author is either a lie or anecdotal.

While these biases are important to keep in mind, they may not matter very much in this study as it essentially found a negative result. The study reports that their intervention (brain stimulation) had no effect on rate of learning or performance on a plane simulation.

Study Design:

Transcranial direct current stimulation (or tDCS) is an old technique but a newly popular way to stimulate the brain without invasive surgery. “Transcranial” means through the skull and “direct current stimulation” means it uses direct current (rather than alternating current) to stimulate the brain. The technique itself is quite controversial in what it can do. Two recent meta-reviews have concluded that single sessions of tDCS have no effect on healthy subjects (Horvath, Forte, & Carter, 2015a) and tDCS has very little effect on the brain’s physiology (Horvath, Forte, & Carter, 2015b). That being said, there are many studies that claim tDCS does have effects in healthy subjects for things like working memory tasks, verbal memory tasks, and motor skill learning (in the elderly). For a review, see Coffman, Clark, Parasuraman, 2014. (Note: Brian Coffman is an author on the paper in question as well)

An enhancement of memory and motor skills would be very useful in a variety of situations, including training airplane pilots. HRL Laboratories set out to test this, presumably for Boeing who is a major manufacturer of airplanes. However, the authors declare no financial conflict of interest despite working for a company owned by Boeing. The authors also do not declare where the funding for the study came from. This disclosure would be standard almost all scientific articles.

The authors used 32 right handed male employees of HRL Laboratories as study participants and assigned each to one of four groups; dorsolateral prefrontal cortex (DLPFC) stimulation, DLPFC sham, primary motor cortex (M1) stimulation, and M1 sham. The sham groups had the tDCS electrodes placed on their scalps and did the study just the same as the stimulation group but the simulator was never turned on. Despite my reservations about this study (small sample size, company employees) this is a great study design.

On top of tDCS (or sham), the participants also had their brain activity monitored via EEG (monitoring brain waves) and fNIRS (a light weight version of fMRI which measures blood flow, and presumably energy demands stemming from increases activity, in the brain). These are awesome techniques to use in conjunction with tDCS to observe if the stimulation is having an effect on the physiology of the brain.

The study used a flight simulator task (seen below) as well as other tasks to study working memory (N-back), attention (situational awareness task) and motor skills (finger-tap task). I will focus on the flight simulator for brevity and relevance to the misconstrued news articles.

Figure 1 Flight simulator 2016 Choe, Coffman, Bergstedt
Figure 1 © 2016 Choe, Coffman, Bergstedt, Ziegler and Phillips. Shared in accordance with CC license 4.0

 

Participants were tested in the flight simulator on their landing skills for 5 trials each day for four consecutive days. This allowed the rate of learning and performance to be measured across days. After getting a resting EEG recording, the tDCS was turned on (or kept off for the sham group). First, participants were trained on how to use the flight simulator by doing some basic training exercises. Participants watched the autopilot land the plane then had to land the plane themselves. Performance was measured on how many G’s were produced, their deviation from the flight path, and vertical speed of the airplane. After the easy landing (during day time), each participant did a nighttime landing. The tDCS was shut off and participants had to do exceedingly harder landing paradigms; nighttime with no runway lights, a mountain landing, and a turbulent landing.

Where are the trained pilots?

You may be wondering where the trained pilots come in to the study and what their brain activity was used for. After all, the press release has quoted Dr. Matthew Phillips to saying,

“We measured the brain activity patterns of six commercial and military pilots, and then transmitted these patterns into novice subjects as they learned to pilot an airplane in a realistic flight simulator”.

Well, if they did measure the brain activity of commercial and military pilots, they totally forgot to mention it in their journal article. Furthermore, their stimulation protocol, as far as they report, was not “transmitting patterns to novice subjects.” Rather, they just turned on the stimulator. There was no special pattern of stimulation. There was no trained pilot derived stimulation pattern. It was just on or off. If you are feeling sanguine, we can give them the benefit of a doubt that they did use trained pilots, maybe to help narrow down which brain regions may be important for piloting an aircraft, at some point and just did not report it in their paper.

They did take great care in their electrode placements and used modeling software to estimate how the current of the multiple electrodes would flow through the scalp, skull, and brain.

Results:

DLPFC:

Comparing the results of participants in the DLPFC stimulation group to sham stimulation group, no learning differences were observed. Directly from the paper:

“Meta-learning regressions did not show statistically significant changes in learning rates between stim and sham groups.” Further, “There were also no statistically significant differences in offline learning rates” Offline learning is the learning that takes place between trials, in this case, the first trial of a given day versus the last trial of that day.

However, for the DLPFC groups, stimulation did reduce the variability in performance (G-force) between days 3 and 4 (seen below, Figure 5A and 5B). That is to say participants in the stimulation group have very similar performances (small, tight deviation lines seen on trials 10-15 and 15-20, Fig 5A) compared to the sham group who had quite varied performances (Fig 5B). However, as stated above, their average performance was the same. Missed landings between the two groups were also not different.

It is interesting to note, in the stimulation group, there was a decrease in oxygenated blood flow and increase in theta band power. A decrease in blood flow is a good thing. This means the brain is using less energy and, since performance was no different, the brain was being more efficient. It had the same performance with less energy when the DLPFC was stimulated.

 

Figure 5 © 2016 Choe, Coffman, Bergstedt, Ziegler and Phillips. Shared in accordance with CC license 4.0
Figure 5 © 2016 Choe, Coffman, Bergstedt, Ziegler and Phillips. Shared in accordance with CC license 4.0

 

 

M1:

Comparing the M1 stimulation group to the M1 sham group, no difference was found in learning or performance, including performance variability. These was also no difference in missed landings.

If you look at the graph 5D, the motor sham group actually had performance variation levels similar to the DLPFC stimulation group. This may indicate the DLPFC stim group’s variation is not due to the stimulation and may rather be caused by other factors.

Conclusions:

A bit boring relative to the Matrix-related press release, huh? Nothing really happened on the learning side or for performance, aside from a reduction in variability for performance in the DLPFC stim group. And that is totally fine. Findings like these are extremely important especially for hot and emerging technologies like tDCS. Further, the authors in their paper absolutely acknowledge these limitations, for example in their “Future Direction” section they state:

“While our results show decreased variability in training, it is too early to confirm or deny any useful improvements to simulation training until an understanding of the sources and contributing factors to the observed behavioral variance is achieved”.

Their neurophysiological findings are quite interesting. Even if DLPFC stimulation does not help us learn quicker or perform better, it may aid our brains in processing information more efficiently (as measured by Hboxy,Hbtot aka blood flow). This is a very cool result in itself.

The authors do point out in their article that a big problem is how variable performance can be. As noted above, this is a big problem with small sized studies like this one. People are very different in how they learn and in the background of experiences they bring to an experiment. Perhaps some of the participants played video games and had no problem with all the controls and thus learned much quicker than people who first had to get comfortable with the controls.

So no, there was no skill uploading like the Matrix, as the press release alludes to. As far as their article goes, there weren’t even any trained pilots to upload memories from. However, their reported results do have merit and are quite interesting. Hopefully HRL Laboratories explores these findings further with more participants and provides the public with more transparent press releases.

This article has been updated on March 10th, 2016 to include a response by HRL Laboratories Media Services
This article has been updated on March 13th, 2016 to include a response by the Journal Frontiers

References:

Choe J, Coffman BA, Bergstedt DT, Ziegler MD, and Phillips ME. 2016. Transcranial Direct Current Stimulation Modulates Neuronal Activity and Learning in Pilot Training. Front. Hum. Neurosci, 10. http://dx.doi.org/10.3389/fnhum.2016.00034

Horvath JC, Forte JD, and Carter O. 2015. Quantitative Review Finds No Evidence of Cognitive Effects in Healthy Populations From Single-session Transcranial Direct Current Stimulation (tDCS). Brain Stimulation, 8(3) 535-550. doi:10.1016/j.brs.2015.01.400

Horvath JC, Forte JD, and Carter O. 2015. Evidence that transcranial direct current stimulation (tDCS) generates little-to-no reliable neurophysiologic effect beyond MEP amplitude modulation in healthy human subjects: A systematic review. Neuropsychologia, 66, 213-236. doi:10.1016/j.neuropsychologia.2014.11.021

Coffman BA, Clark VP, and Parasuraman R. 2014. Battery powered thought: Enhancement of attention, learning, and memory in healthy adults using transcranial direct current stimulation. Neuro-enchantment, 85(3) 895-908. doi:10.1016/j.neuroimage.2013.07.083


From: Blake Porter <blakeporterneuro@gmail.com>
Date: Monday, March 7, 2016 at 6:47 PM
To: Michele Durant <media@hrl.com>, “Phillips, Matthew E.” <mephillips@hrl.com>
Subject: Questions regarding tDCS brain stimulation paper

Hello HRL and Dr. Phillips,

I am a PhD candidate at the University of Otago in New Zealand and have a passion for science communication. I was struck by articles on various news outlets referring to your recent Frontiers in Human Neuroscience article on tDCS. The topic of implanting knowledge/memories via tDCS piqued my interest, especially with recent optogentic papers in mice showing memory implantation. However, I was very disappointed and shocked when I read your article as there was no reports of any trained pilots nor a special tDCS stimulation protocol.

I was wondering if you could clear some things up in regards to this. Was their data on trained pilots that did not make it into the final paper? Did you use some sort of special stimulation protocol for deliver your tDCS that was unreported as well? I saw you took great care in your electrode placement and current levels, but did you have some sort of programmed stimulation pattern derived from trained pilot brain activity as reported in the press release?

I have a commentary piece on my blog regarding your article which can be found here: http://www.blakeporterneuro.com/no-we-did-not-just-upload-memories-matrix-style/

If you would like to comment on my above questions or anything else in my blog post I would greatly appreciate it and will immediately include it in my blog post if you would like me.

Aside from the press release discrepancies, I thoroughly enjoyed your paper and the use of EEG and fNIRS in conjunction with tDCS. I look forward to more of your research especially, as you point out, a study with more participants and perhaps different learning rate groups would be extremely useful.

I look forward to hearing from you,

Blake Porter
PhD Candidate
Neural Mechanisms of Self-control
University of Otago,

Dunedin, NZ

On Wed, Mar 9, 2016 at 2:44 PM, HRL Media Services <media@hrl.com> wrote:

Blake
The confusion stems from the fact that the results from the expert brain patterns are the subject of subsequent manuscript(s). This fact was not included in our original press release by error and it has been corrected. http://www.hrl.com/news/2016/0210/
Our press release never made the claim that we can upload any information to anyone, but that we discovered that low-current electrical brain stimulation can modulate the learning of complex real-world skills.
We appreciate your feedback and hope you will stay tuned for our further developments.
Best Regards,
Media Services
Hrl Laboratories
From: Blake Porter <blakeporterneuro@gmail.com>
Date: Wed, Mar 9, 2016 at 4:00 PM
Subject: Re: Questions regarding tDCS brain stimulation paper
To: HRL Media Services <media@hrl.com>
Cc: “Phillips, Matthew E.” <mephillips@hrl.com>

Hi HRL,

Thank you for getting back to me and so quickly, I greatly appreciate it.

I do agree with you the press release did not directly use “upload” but it does prime the reader with “imitating” the Matrix followed by Dr. Phillips’ quote that they, “transmitted these patterns into novice subjects” It is quite a small and obvious logical leap to take this as “uploading”. To upload something is to transfer it from one place to another. Arguing you didn’t us the word upload is just a weak argument of semantics. Indeed, in the Matrix information (such as expert pilot knowledge of a helicopter) is transmitted over a cable to those plugged in. Your press release claims information (expert pilot brain activity) was transmitted/transferred/uploaded to naive subjects. Whether you used the word upload or not the concept of transferring information from one place to another (AKA uploading) is clearly implied.

What I would really like to know is if the tDCS protocol you used had some sort of pattern to it derived from the expert pilots. As I am sure you are aware, direct current has no patterns; that is why it is direct current and not pulsating direct current or variable direct current. If there was some sort of pattern to the stimulation this would need to be reported in the journal article, which it is not.

Further, I am still perplexed at your press release claims that tDCS “improved their piloting abilities.” and that it was “effective in accelerating practical learning”. Your labs own paper contradicts these very claims,
“Meta-learning regressions did not show statistically significant changes in learning rates between stim and sham groups.” Further, “There were also no statistically significant differences in offline learning rates” Finally, “The number of missed landings (did not land before or during the runway) was not different across groups over days”

How could you possibly “accelerate” learning and at the same time have no change in learning rate?

Looking deeper into your supplemental, it is clear why you only put in G-force because all other performance measures couldn’t even make the weak variability claim. How could abilities (plural) have been improved when only one performance metric showed any modulation and that modulation was not even an improvement in performance?
The video press release is even more outrageous claiming you could “take a group of individuals and train them to a similar level than we could without the brain stimulation”. There was no data on expert pilots even reported and stim and sham groups had the same performance levels.

I, and other scientists and clinicians*, would greatly appreciate the press release be revised to a degree to reflect the paper and clearly state no expert pilot information was transmitted or even used in this study.

Blake

*In case you are unaware, I am not the only one disappointed by your false claims:
Academic Neurologist Dr. Steven Novella: http://www.theskepticsguide.org/podcast/sgu/556

 


Creative Commons License
This work by Blake Porter is licensed under a Creative Commons Attribution-Non Commercial-ShareAlike 4.0 International License


If you enjoyed this article and found it informative, consider donating! Funds first go to keeping the website running (domain name/hosting) and secondly go to me. Donations are secure and handled by PayPal.




Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.