In the next 6 to 8 months, we are likely to see significant advancements in AI.
A new generation of models is expected to be released and trained using 5 to 10 times more computing power than current models like GPT-4.
The key point to note is that between March 2023 and now, all models like GPT-4-Turbo, GPT-4o, and GPT-4-mini were outcomes of additional post-training, essentially pulling to the same model we’ve had for 18 months.
The same applies to improvements like Claude 3 to Claude 3.5 and Gemini 1.5 iterations.
𝙄𝙣 𝙩𝙚𝙧𝙢𝙨 𝙤𝙛 𝙩𝙞𝙢𝙞𝙣𝙜, we 𝙜𝙪𝙚𝙨𝙨 𝙂𝙋𝙏-5 𝙬𝙞𝙡𝙡 𝙗𝙚 𝙩𝙝𝙚 𝙛𝙞𝙧𝙨𝙩 𝙩𝙤 𝙧𝙚𝙡𝙚𝙖𝙨𝙚 𝙞𝙩𝙨 𝙣𝙚𝙭𝙩-𝙜𝙚𝙣 𝙢𝙤𝙙𝙚𝙡𝙨 (𝙥𝙧𝙤𝙗𝙖𝙗𝙡𝙮 𝘿𝙚𝙘 𝙩𝙤 𝙁𝙚𝙗), 𝙖𝙨 𝙊𝙥𝙚𝙣𝘼𝙄 𝙞𝙨 𝙨𝙡𝙞𝙜𝙝𝙩𝙡𝙮 𝙖𝙝𝙚𝙖𝙙 𝙤𝙛 𝙤𝙩𝙝𝙚𝙧𝙨.
𝘈𝘵 𝘵𝘩𝘦 𝘴𝘢𝘮𝘦 𝘵𝘪𝘮𝘦, we 𝘵𝘩𝘪𝘯𝘬 𝘊𝘭𝘢𝘶𝘥𝘦-4 𝘰𝘳 𝘎𝘳𝘰𝘬-3(𝘸𝘩𝘪𝘤𝘩 𝘌𝘭𝘰𝘯 𝘴𝘵𝘢𝘵𝘦𝘥 𝘸𝘪𝘭𝘭 𝘣𝘦 𝘢𝘷𝘢𝘪𝘭𝘢𝘣𝘭𝘦 𝘪𝘯 𝘋𝘦𝘤𝘦𝘮𝘣𝘦𝘳), 𝘸𝘪𝘵𝘩 𝘓𝘭𝘢𝘮𝘢-4 𝘯𝘰𝘵 𝘵𝘰𝘰 𝘧𝘢𝘳 𝘣𝘦𝘩𝘪𝘯𝘥. (𝘋𝘰𝘯’𝘵 𝘵𝘢𝘬𝘦 𝘪𝘵 𝘵𝘰𝘰 𝘴𝘦𝘳𝘪𝘰𝘶𝘴𝘭𝘺.)
It’s hard to say exactly what the new models will be capable of right now, but it will be interesting to see how quickly AI improves in the coming months.
