Have you heard the buzz about large language models (LLMs) such as ChatGPT? They’re programming wonders that can instantly pen essays, brainstorm meals, or even help craft your emails. But as amazing as they are, they’ve not been historically great at challenging tasks like problem-solving, especially when it comes to mathematics or complex reasoning. However, this shortcoming is becoming less and less of an issue.
Nun hält eine neue Generation von LLMs Einzug, die als „Reasoning-Modelle“ bezeichnet werden und bei der Bewältigung komplexer Aufgaben erhebliche Fortschritte zeigen. Im Gegensatz zu ihren Vorgängern, die sich stark auf Sprachmuster stützten, um Antworten zu erraten, wenden diese Modelle bewusstere, schrittweise Strategien an, ähnlich wie es ein Mensch tun würde.
What’s more, researchers at MIT’s McGovern Institute for Brain Research noted a striking similarity between how humans and these new models approach difficult tasks. Interestingly, they discovered that tasks which demand the most mental effort from humans are also the ones that require the most computational strain from reasoning models. This lead to a new concept: the “cost of thinking” transcends the human-machine divide.
Diese etwas unerwartete Konstellation führte das MIT-Team unter der Leitung von Associate Professor Evelina Fedorenko überraschenderweise. Die Entwickler dieser Modelle konzentrieren sich in der Regel darauf, ein System zu schaffen, das unter zahlreichen Bedingungen gut funktioniert und genaue Ergebnisse liefert, anstatt das kognitive Verhalten des Menschen nachzuahmen. Daher war die Konvergenz von menschlicher und maschineller Leistung eine unerwartete, aber spannende Entdeckung.
These reasoning models are still fundamentally artificial neural networks – systems that learn by analysing data and recognizing patterns. However, they go a step further than their predecessors by addressing deeper cognitive tasks, such as math problems or coding. A key innovation is their approach to problem-solving: these models break problems down into smaller parts, which considerably enhances their performance.
Ingenieure nutzen zudem das verstärkende Lernen, um diese Modelle zu trainieren, wobei richtige Antworten belohnt und falsche bestraft werden. Das Modell lernt im Laufe der Zeit, Lösungswege zu erkunden, die häufiger zu korrekten Schlussfolgerungen führen, und ahmt so einen eher menschenähnlichen kognitiven Prozess nach. Diese traditionelle Methode dauert zwar länger als die von früheren LLMs verwendeten Verfahren, verbessert jedoch die Genauigkeit erheblich.
Andrea Gregor de Varda, Postdoktorandin am MIT K. Lisa Yang, ICoN-Zentrum, along with Fedorenko, conducted an experiment to determine this theory. They not only observed the accuracy but also gauged how much effort was required. For humans, this involved tracking response times to the millisecond. For models, they looked at how many tokens, or internal pieces of language, the model generates while working through a problem. Apparently, the harder the problem, the more tokens the model generates – much like how we humans metaphorically ‘talk to ourselves’ when encountering a tricky problem.
Seven types of problems, including arithmetic and intuitive reasoning, were posed to both humans and the reasoning model. Expectedly, harder problems took longer for humans to solve and also required the reasoning model to produce more tokens. However, while these findings are compelling, de Varda cautions against jumping to the conclusion that these models fully mirror human cognition. He highlights that as they still function primarily in an abstract, non-linguistic space, there’s still more to learn about how closely they model human thought processes.
Many questions remain unanswered. For instance, do these models represent information like our brains do? Can they tackle issues requiring real-world knowledge beyond their training data? As researchers explore these frontiers, one intriguing suggestion is clear: machines might slowly but surely be evolving closer to human-like cognition, not because they were explicitly programmed to, but possibly because it’s simply the most effective way to think.
Erfahren Sie mehr über die komplexe Beziehung zwischen menschlicher und maschineller Kognition im vollständigen Artikel unter MIT-Nachrichten.
Diese Website verwendet Cookies.