The laborious delivery of Markram’s brainchild

The European multimillion research project Human Brain Project is under fresh leadership, its visionary founder Henry Markram ousted from power. Worse, both life-and-blood projects of the neuroscience professor at the Swiss elite university EPFL are not performing as he envisioned them to.  His open access publishing house Frontiers was dumped by the Nature Publishing Group and ended up with the German publisher Holtzbrinck, with neither them nor Frontiers particularly keen on boasting this property relationship. Editors and reviewers ran away or refused duty in droves, Frontiers was even fingered as a possible predatory publisher. On top of everything, Markram’s outlet stands accused of being an employee-abusing spam factory.

So much for Markram the Entrepreneur and Inventor. But he is first and foremost a scientist, whose research proposal received the biggest research funding grant in history: one billion Euros from the European Union, for his “brainchild” (as journalists dubbed it), the Human Brain Project. The modest promise Markram originally made to secure this mind-boggling mountain of cash: he intended to simulate the entire human brain in his supercomputer by 2023, the possibility of artificial consciousness specifically not excluded. Now however, his consortium partners took over, Markram was dethroned in a scientists’ coup and pushed aside to tinker on his seemingly less ambitious, but just as science-fictionary mouse Blue Brain simulation. Once in control of almost everything and everyone, with all the big money going through his hands, Markram is now only one of 12 project leaders and far from being the boss. The Human Brain Project (HBP) became instead a kind of funding network without any properly defined orientation, not even the new leaders could convincingly describe any defined goals. Instead, its main purpose seems to be now set on keeping the EU funding of almost €50 Million per year flowing. Remarkably, all this was achieved after an allegedly independent mediation by the director of an HBP-consortium partner institute; coincidentally a member scientist from this Jülich Research Centre (Forschungszentrum Jülich, located in a rural triangle between German cities Düsseldorf, Cologne and Aachen) is now the new scientific director of HBP.

This is how the dream of the brain-in-the-box grew, prospered and imploded.

Ideas Worth Shredding

In 2009, Markram gave a TED-talk, titled modestly “A Brain in a Supercomputer”, which he opened with: “Our mission is to build a detailed, realistic computer model of the human brain”.

The talk and its scientifically unfounded claims, outrageous promises and gratuitous self-promotion will probably hurt your own brain. To Markram, building a brain is only a question of computational power and is all about linearly upscaling a simulation of one single cortical column (something he already in 2012 claimed to have achieved, with many of his peers disagreeing that mapping a structure is by far not the same as simulating a function). Regardless of the scepticism, Markram concluded his 2009 TED speech with:

“I hope that you are at least partly convinced that it is not impossible to build a brain. We can do it within 10 years, and if we do succeed, we will send to TED, in 10 years, a hologram to talk to you”.

There are 3 more years to wait till that prophesied TED visitation, but it is unlikely that any artificial intelligence hologram will be coming. Markram specifically nourished and never excluded the possibility that his creation will develop consciousness. Even in 2013, he cautiously answered to The Guardian: ”That we are not sure about”.

The Funding Freakshow

Back in 2011, 6 Flagship projects were competing to receive €1 Billion funding, one was weirder than another (cushy robot companions for lonely people, planetary-scale human civilisation simulators and nano-sensors which would warn you of toxic substances in your environment).  HBP was eventually chosen by the EU in early 2013, together with another mega-project, Graphene, dealing with new type of super-conductor technology. The HBP report to the European Commission from 2012 (available here) sounded like a work of science fiction. It declared:

“The HBP High Performance Computing  Platform  should  provide  the  project  and  the community with the computing power it needs to build and simulate models of the brain”.

HBP promised to generate realistic computational models of human and animal behaviour, consciousness, language and simulate any brain disorder while designing personalised medicines. One goal was to study “integrative principles of cognition”:

“Researchers should use the Brain Simulation and Neurorobotics platforms in projects that systematically dissect the neuronal circuits responsible for specific behaviours, simulating the effects of genetic defects, lesions, and loss of cells at different levels of brain organisation, etc and modelling the effects of drugs. The ultimate goal should be to model the unique capabilities that distinguish humans from other animals, in particular language”.

This slideshow requires JavaScript.

Figures from HBP report to EU, 2012

That proposal was however heavily modified before it was finally approved by the EU, the approved version is not available to outsiders. Yet most of that unscientific nonsense was used on the HBP website between 2012 and 2014, as my backup shows. This quote from the 2012 HBP website gives a fascinating insight, not really into brain science, but into Markram’s own narcissistic brain:

“The Blue Brain Project has already shown the feasibility of the strategies proposed by its successor. Between 2005 and 2011, Henry Markram’s EPFL team developed facility with the tools, the know­how and the supercomputing technology necessary to build models of the brain. This facility will provide a starting point for the development of the much larger facility planned in the Human Brain Project. This will be developed according to a well­planned roadmap. First, it will develop the capability to build models brain regions (“mesocircuits”), then of brain systems (“macrocircuits”) and finally of complete brains. Before attempting to simulate the complete human brain, it will create the tools and knowhow to model simpler brains such as those of mice, rats, cats, and monkeys. As the project progresses the facility will develop the knowhow and tools to model and simulate the brain of any animal, at any stage of its development, in any state of health or with any specific disease”.

All For a Good Cause

Both flagship projects, HBP and Graphene originally assumed to receive each €1 Billion for ten years, no questions asked. However, already by the end of 2012 the EU Commission declared that they only meant to give maximum half a billion, the other half was to come from EU member states and the participating research networks themselves. The EU soon became even more cautious and moved to dispense HBP funding in 2 year increments, calculated separately each time. From 2014 on, the first two years delivered “only” €54 Mio, the next two years €89 Mio and for 2017/2018 HBP was assigned another €88 Mio. Which makes it theoretically possible for HBP to collect the bulk or even all of the earmarked half a billion Euro till 2013, without actually delivering what they so boldly promised, namely a human brain in silico. A standard occurrence in research funding unfortunately, it is only the outrageous amount of public funding invested which makes the HBP charade so outstanding.

Why did EU politicians and bureaucrats fall for such naive science fiction as HBP in the first place? There was one important promise Markram made which everyone could get behind: abolishing invasive animal experiments, and in neuroscience this often means the ethically highly sensitive issue of non-human primates. Not only basic science, also future human medicine development was to be done using his in silico models of human brains (plus those of “mice, rats, cats, and monkeys”, probably also badgers and giraffes as well, time permitting). Reducing animal experimenting is indeed a very noble goal, and one can understand (up to a degree) why so many became excited by the unscientific bigmouth proposal. If nothing else, the lack of sufficient life science education among our politicians, bureaucrats and unfortunately also science journalists is to blame for this. Markram declared in his TED-Talk:

“We cannot keep doing animal experimentation forever, and we have to embody all our data and all our knowledge into a working model. It’s like a Noah’s Ark”.

Later on, when HBP criticisms became serious, the Swiss visionary insisted he never said that his virtual brain would make animal experiments obsolete.

Save Our Funding!

Markram felt too secure in his power position and ruled over HBP like an absolute monarch. He was only supervised by himself, controlled all money flows (often towards his own and his close associates’ research) and even sacked an entire network of 18 laboratories out of the consortium. Eventually, the protests of his peers became public. Over 800 scientists signed in summer 2014 an “Open message to the European Commission concerning the Human Brain Project”, asking that at the impending HBP review

“the European Commission must take a very careful look at both the science and the management of the HBP before it is renewed”.

The signatories, all of them European neuroscientists, also demanded:

“In the case that the review is not able to secure these objectives, we call for the European Commission and Member States to reallocate the funding currently allocated to the HBP core and partnering projects to broad neuroscience-directed funding to meet the original goals of the HBP—understanding brain function and its effect on society”.

HBP was therefore in direct danger to be abolished and its half a billion Euros redistributed among the European scientific community, all because of Markram’s hubris. Something had to be urgently done to prevent the obscenely huge HBP budget from slipping away, hence a mediation committee was formed. The European Union described it in a later announcement as “independent”, others news sources wisely avoided using this word. The invited mediator namely came from a key HBP partner and its central provider of the super-computer power, the Forschungzentrum Jülich (FZJ) in Germany. Wolfgang Marquardt was himself not a member of HBP, but as director of his research institute he is of course always evaluated by the amount of external funding he manages to recruit to it. In any case, Marquardt represents the major EU research donor Germany in the HBP stakeholder board, which acts as there as “Ultimate Decision-Making Body”. The mediating committee debated over the winter of 2014/2015 and consisted of HBP members (sans Markram) and their critics (though not the harshest ones). Marquardt explained to me now:

“A potential conflict of interest was declared by me when the first mediation request came (in August 2014). Even though I am responsible for the entire FZJ, I had no function in a panel of HBP […] and I had therefore no direct CoI. Besides, due to my past activities I also felt myself very well qualified to successfully carry out the mediation process and to separate between my role as CEO [of FZJ, -LS] and as mediator. It was also difficult to actually find a person who would have been qualified to take over the mediation and had no relation to the HBP and its environment. An agreement on the implementation of the mediation process has been made between me personally and the BoD [Board of directors, -LS]”.

Marquardt negotiated a compromise, which can be briefly summarised as: switch from whole-brain simulations to humbler mapping, Markram and his brain-in-the-box demoted, Markram-obedient Board of Directors abolished and money distributed more equally, away from EPFL and towards other partners (for details, see full report of the mediation committee from March 9^th 2015).

Magic Cell

However, more was needed to convince EU to continue paying. After 3 years and almost €150 Mio spent, HBP delivered no published results worth mentioning. But then, none other but Markram came as the rescuer to all those who deposed him. He published a paper in the elite journal Cell (Markram et al, 2015), which is to go in history maybe as not the most scientifically useful research paper, but certainly as the most money-bringing one. Just after the paper appeared, the EU Commission lauded this “remarkable work” about “a first draft digital reconstruction of the micro-circuitry of a section the cortex of a rat brain” and announced their intent to “support the HBP until 2020, under the EU research and innovation programme Horizon 2020”. This in turn means: many individual grant applications to Horizon2020 will go empty because an enormous yet undisclosed multi-million sum has now been allocated for HBP.

So what did Markram discover in his Cell paper, which opened up the EU Sesame? First of all, it is not a research, but a “resource” paper, meaning its scientific novelty or originality is rather limited. David Hansel, research director at Université Paris Descartes in Paris, is a long-standing critic of HBP and its philosophy. To me, he pointed out that the HBP predecessor Blue Brain Project “did not lead to any notable  breakthrough that would be specific to its specific objectives” and described the one-billion Euro heavy follow-up HBP as “totally crazy and highly dishonest, without scientific basis, especially since Markram claimed that this will end animal experiments”.

In regard to Markram’s highly lauded Cell “resource” study, Hansel (who specialises in visual cortex and large neuronal network modelling) provided this expert opinion:

“The theoretical modelling aspect is nonsense and offers no truly new results. Most of these simulation analyses can be done more simply, on a home laptop. The authors were unable to provide any new results specific to their new approach. What is declared as major advances is either not proven or is already known, from simpler models. The claim that they created a database to be used for “future generations” [as declared in a press release by the paper’s senior author and dedicated Markram follower Idan Segev, -LS] is exaggerated since these techniques are developing at a very rapid pace”.

Funnily, HBP originally promised to EU that “as far as possible, papers should be published in Open Access Journals and/or deposited on pre-print servers”. For that Cell paper, Markram et al did neither. Whoever wants to read it has to pay Cell $31.50. Instead of a preprint, there was only an unsuccessful previous attempt at Nature, where the paper was rejected after a devastating peer review.

The King is Dead, Long Live the King!

HBP has completed its re-organisation just recently. Even the website changed radically, old Markram-era documents are still searchable, but cannot be accessed without a login anymore. The coordination and the financial management of the Consortium remains at EPFL, Director General is still the EPFL provost Philippe Gillet. With the Board of Directors dissolved, Markram became just one of 12 sub-project leaders on the Science and Infrastructure Board. The times when he and two other members of the executive committee steered the HBP flagship as triumvirate, are over. Richard Frackowiak, clinical neurologist at the University of Lausanne (right next to EPFL) resigned completely from the HBP and Karlheinz Meier, physics professor at University of Heidelberg and leader of Neuromorphic Computing subproject at HBP was “demoted” to supervise the infrastructure.

One of former Board of Directors members has recently been elected to preside over the Science and Infrastructure Board. Coincidence or not, after a mediation by FZJ-director and HBP stakeholder board member Marquardt, this new scientific director is Katrin Amunts, head of the FZJ-department of Neuroscience and Medicine (she also leads a department at the University Clinic Düsseldorf). Amunts’ research interests: to build a “3-D model of the human brain which considers cortical architecture, connectivity, genetics and function”. She now explained me what exactly HBP intends to deliver under her scientific leadership:

“The objective of the HBP is to build an ICT [information and communication technology, -LS] -based infrastructure in Europe, which will enable researchers to create insights in neuroscience, computing and brain diseases. This is designed to help better understand the principles of organization of the human brain – as a matter of basic science, medicine, or in the sense of humanity sciences (for example, what makes us human?). Simulation is an important goal, but at the same time also an instrument to ask new questions and to verify experiments. The project should also lead to development of new technologies and methods – for example, in the area of neuromorphic computing, HPC [High-performance computing, -LS], or robotics or data analysis, for example, in the area of Big Data“.

I still do not understand what exactly the new overarching scientific goal of the HBP is and why they need half a billion Euros for it (maybe my readers can contribute with a discussion). To me, it looks like a general computational neuroscience funding network inside the wider Horizon 2020 funding network. The information below confirms my suspicion that the main focus of HBP is that of securing money, in the reasonable expectation that the EU will wish to avoid losing face and continue paying its Flagship to travel to whatever destination the wind blows it, until 2023.

Amunts explained to me that the HBP funding is set by the Framework Partnership Agreement from October 2015 and decided in consecutive two-year agreements, evaluated by an independent international reviewer panel. Each bi-annual amount of funding money is set by Specific Grant Agreements (SGA); HBP participants also hope for equivalent funding from member states and other sources. It is apparently a constant haggling and begging, with many arms to twist for a hand-out. Amunts also mentioned that the EU-funded FLAG-ERA (which in turn gives money to the two flagships HBP and Graphene) directs new partners with their own funding to HBP. There are also discussions with the US-counterpart of the EU flagship, the BRAIN Initiative, to join forces in the common quest for funding.

Also “outsider” participants can join HBP, the application windows however open every 2 years, when the SGAs are negotiated. After the French cognitive psychologist Stanislas Dehaene and his colleagues (whose Cognitive Architectures section Markram first sacked, then re-instated under mediation pressure) decided last summer to depart from the Consortium anyway, HBP used the new bi-annual SGA to recruit new participants. According to Amunts the new branch of cognitive and systems neuroscience was recruited through external evaluation and will receive 10% of the budget.

Epilogue

The fallen HBP patriarch Markram was recruited to EPFL in 2002 by its then freshly appointed president Patrick Aebischer. Ever since Aebischer (who prides himself to lead his university like a business) held his protective hand over Markram, and supported his projects with words and funding. With such patronage, most Markram critics knew to keep quiet. Both Swiss federal universities EPFL and ETH pitched together to obtain for Aebischer’s protégée and his mouse-brain simulating Blue Brain project the IBM supercomputer Blue Gene. The EPFL president then played a key strategical role in making the billion-euro HBP possible, while ensuring that the lion’s share of funding money and the controlling power went to his EPFL (Aebischer also now remains the Swiss representative of the HBP stakeholder board). However, the grand man is about to retire by the end of 2016. New EPFL president will be the computational scientist Martin Vetterli, who seems to have no connection to HBP or Markram whatsoever.

Maybe reading his own Frontiers will cheer Markram up. There are some fun papers inside, about paranormal activities, madness-causing bread and, most recently, poisonous “chem-trail” conspiracies.