Yes but also no. There's answers at different depths here.

LLMs are what every normie call "AI", and those obviously don't think. They are essentially the statistical correlate of all text produced by our civilization. Like a playback mechanism for surface-level knowledge. There will be isolated areas where their statistics actually do capture the underlying model and as such are real "knowledge", but we currently cant control this and barely detect it. However, slight magic does happen when they are folded back on themselves, as agents, writing an output, then correlating that with the statistical grand total of human knowledge to create a response, repeatedly. How many "new discoveries" truly aren't the combination of patterns already known from other contexts? Not many.

Then there's other architectures. Go was considered the prime example of games that couldn't be beaten by algorithms, but after AlphaGo, humans will never beat the machine again. Science made no headway with predicting protein folding, until AlphaFold swooped in and did it (something so much more significant than fcking LLMs). Those weren't solved by processes analogous to human thought, but to call it fake thinking would be not entirely different from accusing airplanes of fake flight.

Various forms of fake thinking will continue to do increasing amounts and types of real work.