An Uncommon Mentor
“Are you coming to the meeting?” he asked, following my answer with “Look, there is no ‘not sure’ here; it’s either yes or no.” This was my first exchange with Gadi Geiger upon joining Tomaso Poggio’s lab (“An Uncommon Education,” January/February 2017). Gadi, as we all call him, made espresso, and we soon started to chat about human vision and hearing and, as he put it, the need for “a new kind of mathematics” for studying the brain. In the Poggio Lab and the Center for Brains, Minds, and Machines, we are studying computational aspects of learning and intelligence and models of sensory processing in the cortex. But Gadi is always skeptical about computation and insists there can be no modeling without constraints from physiology, no simulation results without psychophysics, and no computational cognitive science without neuroscience.
Gadi’s research and mind-set have always involved light, from a physical, biophysical, or psychophysical perspective, as in optics, the visual system, or visual perception. He went from designing lasers to studying, with Poggio, the neural mechanisms responsible for visual orientation in the fly to measuring, with Jerome Lettvin, central and peripheral vision in human visual perception. Through strong intuition, an unconventional perspective, and chance, Gadi and Lettvin approached dyslexia, conventionally related to phonemic awareness or language, as a visual condition. This work generated new perspectives and practices, many of which came from studies and trials with dyslexic children in schools in Germany, Italy, and the United States. The result was new reading paradigms for people with dyslexia and new ways of diagnosing the condition.
At MIT, Gadi worked in the legendary Building 20, which he always recalls with a certain level of nostalgia, alongside Lettvin, Noam Chomsky, and Patrick Winston, among others. He always speaks highly of the unconventional thinking and groundbreaking work of Norbert Wiener, Walter Pitts, David Marr, and, above all, Lettvin (or “my friend Jerry,” as he likes to say). In the Center for Computational and Biological Learning, directed by Poggio, a friend and collaborator from their days at the Max-Planck Institute, Gadi became a mentor and a friend to many students who are now well-established names in neuroscience, artificial intelligence, and machine learning.
Gadi willingly shares his many stories—enough to write books or make documentaries about—and offers commentaries on art, music (the mathematics in Bach), and current social and political affairs over a dinner he has prepared or while browsing through his record and book collection. For Mr. (not Dr.) Geiger, his “uncommon education,” as the MIT News article described it, might be the point of reference for his unconventional views in research and life. MIT attracts and puts to good use uncommon, unconventional minds and characters like Gadi. And in turn, they push MIT forward.
Center for Brains, Minds, and Machines, MIT
“Clean Fuels from Greenhouse Gas” (January/February 2017) does not mention the energy needed to convert carbon dioxide and water to carbon monoxide and hydrogen. In reality, the total process may not be environmentally friendly unless the energy is produced using the excess capacity of an environmentally friendly power plant. One might also question what the energy efficiency of the process will be, including distribution losses.
Emil M. Friedman ’68
Hillside, New Jersey
Professor Yogesh Surendranath responds:
I concur wholeheartedly with the concern raised by the reader. Carbon dioxide is an inherently energy-poor molecule, so any strategy for upgrading this molecule to a fuel will require an energy input greater than the energy content in the resulting fuel (distribution losses will impose an additional energy penalty). Thus, electrochemical conversion of carbon dioxide to fuels is not a viable strategy for remediating existing carbon emissions if the extra energy input comes from traditional fossil fuels.
However, the process is a potential strategy for storing the energy from renewable sources such as solar and wind by converting it to the energy-dense chemical bonds of a CO2-derived fuel, as occurs when solar energy gets converted to chemical energy in natural photosynthesis. Importantly, the amount of energy needed for carbon dioxide fixation is dramatically increased if the chemical process produces undesirable by-products, which would need to be separated and further processed downstream. In this study, we developed a strategy for tuning the selectivity of the catalyst to avoid the production of undesirable by-products, thereby enhancing the efficiency of the process.
Tell us what you think
Write MIT News, One Main Street, 13th Floor, Cambridge, MA, 02142