Yann LeCun, head of AI at Meta, and Jake Browning argued that 'a system trained on language alone will never approximate human intelligence even if trained from now until the heat death of the universe,' suggesting that language-only training has fundamental limitations.