Scientific Committee

From the Scientific Committee

Mesvera and the Scientific Committee

Like you, we have been watching AI's rise in academia and science with astonishment. We start to admire it—"It can do that, too?"—and then, seeing the methodological, ethical and institutional threats that shake traditional scientific method to its roots, we say "Easy now..." We take a stand against this citation-breaking, hallucinating creature and try to get back to our work: "Keep it away from us." Then, over evening tea, we hear our 12-year-old daughter say, "Dad, did you hear? Cancer cells can be detected in advance with AI." In the morning we sit across a master's thesis written with AI, full of sources and passages that do not exist, and what little sympathy we had for AI fades at once.

In academic conversations among middle-aged—or, more kindly, experienced—colleagues, the topic of the day is that AI hallucinates: it produces data, sources and citations that contradict reality. The reproducibility crisis has only recently begun to be discussed as a methodological problem. In short, this AI subject is far from settled.

In a word, AI dropped a pile of methodological and ethical problems into our laps at once: ghostwriting, data privacy, straw-man papers, review crises, digital inequality. We have a thoroughly brilliant, self-learning, mischievous child who does not merely ignore the rules—it makes them. (Where we come from they call such a child "ekis.") May it grow up with a mother and a father.

The other side of the coin was the claim that AI-supported work could, in a short time, find answers to problems the scientific world has discussed for years—from drug development to chronic disease—problems thought to take years or decades to solve. We said our "ekis" child could not do that; then we remembered what it had done in two years, hit the brakes again, and fell into a deep silence to think.

What has already been achieved

Those who follow the literature—especially developments outside their own field—saw very quickly what AI has accomplished in a not-so-distant past. Let me set out a few of them:

Proteins

AlphaFold

Discovering the 3D structures (folding patterns) of proteins, the building blocks of life, in the lab (X-ray crystallography and the like) can take years. AlphaFold, the AI model developed by Google DeepMind, predicted the 3D structure of nearly every known protein—about 200 million—at atomic accuracy in only a few years. That output has accelerated projects such as malaria-vaccine development, cancer-drug design and plastic-eating enzymes by decades.

Materials science

GNoME

Developing a new battery technology, solar panel or semiconductor used to mean synthesizing trillions of element combinations from the periodic table, one by one, in the lab. AI simulated which new materials might be stable by predicting crystal structures. Google's GNoME AI discovered about 380,000 new stable crystal structures—ten times more than humanity had found to date. Longer-lasting batteries for future electric vehicles and more efficient solar panels are now produced from the specific formulas AI proposes, without blind trial and error in the lab.

Drug discovery

Halicin

A university in the United States screened more than 6,000 molecules with AI and identified a molecule never previously considered an antibiotic: Halicin. When tested in the lab, Halicin killed even the deadliest and most resistant bacteria. AI brought the drug-discovery process down to a matter of days.

Astronomy

Event Horizon

The volume of data from space telescopes (petabytes) is too vast for the human eye or classical software to analyse. When the first photograph of a black hole was taken with the Event Horizon Telescope, AI algorithms were used to fill in missing data and to synchronise the telescopes. Among billions of signals from deep space, humanity can now tell with high accuracy which belong to a planet and which to cosmic noise—thanks to AI.

Archaeology

Herculaneum

Ancient texts such as the Herculaneum papyri, carbonised in the eruption of Vesuvius and liable to turn to ash the moment they are opened, were considered impossible to read. Texts unread for 2,000 years have begun to be read, word by word, with the help of AI.

Can you tell a flowing river "Do not flow"? The workable answer is to channel it so it does not flood.

That view led us to Mesvera

As you read these sentences I know you too see AI as a flowing river: sometimes a flood, sometimes an astonishingly useful irrigation project. In short, can you tell our mischievous/"ekis" child "Don't run," or tell a flowing river "Don't flow"? The workable answer is to steer it so it does not flood. In other words, the prevailing view for now is that the right path is to manage and regulate methodological and ethical concerns on one side, and on the other to integrate AI positively into how universities produce knowledge.

That prevailing view led us to gather, in a single portal, discipline-specific AI tools that can support universities in producing knowledge, taking out patents and making inventions—new ideas, drugs, treatments, molecules and the rest.

Which AI tools should sit in this portal? Which scientific criteria should we use when we put a tool there? How will we transfer data produced by one tool in the same environment? Who will choose these tools, and how? Fortunately we have a settled, classical methodology for this. We set out with 16 professor colleagues working in different disciplines—from anatomy to physics, from cardiology to history—drawn from the health, social, natural and educational sciences. With a qualified scientific committee we began work according to selection criteria.

Faculty members

Fields of science

Health, social, natural and educational sciences

FDA · WHO

Reference institutions

How the tools are chosen

Mesvera's research and development began under these principles: which AI tools to use; why this tool was chosen among dozens; what its hallucination risk is; that output from one tool must transfer instantly to another; that every selected tool must meet the criteria; and that it must be approved by a scientific committee of faculty members with different fields of expertise.

Scientific selection criteria

When dozens of tools exist, the reason a tool is admitted to the portal rests on scientific references and the committee's criteria.

Hallucination risk

Every tool is assessed for the risk of producing data, sources and citations that contradict reality.

Instant transfer between tools

Data produced in one tool must be transferable to another in the same environment.

Cross-disciplinary committee approval

Every selected tool is approved by a scientific committee of faculty members with different fields of expertise.

To your judgement and criticism

The portal therefore includes AI tools carefully selected by the Scientific Committee; recommended by institutions such as the FDA and WHO; or shown by scientific studies to be effective and useful.

With Mesvera, university faculty and researchers will reach the committee-approved AI tools they need while producing and processing scientific knowledge—with a single password, under a single roof. The AI tools in Mesvera are also capable of supporting undergraduate and graduate teaching from medicine to engineering, from law to education.

In short, we present to your judgement and criticism a Mesvera whose tools are chosen on scientific principles, believed to raise universities' capacity to produce knowledge, limited by faculty and discipline, reachable with a single password, and able to pass knowledge developed in one tool to another for further work.

We hope it serves the Turkish university ecosystem well.

On behalf of the Scientific Committee

Prof. Dr. Selçuk Beşir DEMİR

See why this committee matters to Mesvera

Read why institutions choose a portal whose tools are selected and approved by the Scientific Committee on the Why Mesvera? page.

Explore Why Mesvera