GPT-4’s Secret Has Been Revealed
<p>GPT-4 was the most anticipated AI model in history.</p>
<p>Yet when OpenAI released it in March they didn’t tell us anything about its size, data, internal structure, or how they trained and built it. A true black box.</p>
<p>As it turns out, they didn’t conceal those critical details because the model was too innovative or the architecture too moat-y to share. The opposite seems to be true if we’re to believe the latest rumors:</p>
<p>GPT-4 is, technically and scientifically speaking, hardly a breakthrough.</p>
<p>That’s not necessarily bad — GPT-4 is, after all, the best language model in existence — just… somewhat underwhelming. Not what people were expecting after a 3-year wait.</p>
<p>This news, yet to be officially confirmed, reveals key insights about GPT-4 and OpenAI and raises questions about AI’s true state-of-the-art — and its future.</p>
<h1>GPT-4: A mixture of smaller models</h1>
<p>On June 20th, <a href="https://twitter.com/swyx/status/1671272883379908608" rel="noopener ugc nofollow" target="_blank">George Hotz</a>, founder of self-driving startup Comma.ai leaked that GPT-4 isn’t a single monolithic dense model (like GPT-3 and GPT-3.5) but a mixture of 8 x 220-billion-parameter models. Later that day, <a href="https://twitter.com/soumithchintala/status/1671267150101721090" rel="noopener ugc nofollow" target="_blank">Soumith Chintala</a>, co-founder of PyTorch at Meta, reaffirmed the leak. Just the day before, <a href="https://twitter.com/MParakhin/status/1670666605427298304" rel="noopener ugc nofollow" target="_blank">Mikhail Parakhin</a>, Microsoft Bing AI lead, had also hinted at this.</p>
<p><a href="https://albertoromgar.medium.com/gpt-4s-secret-has-been-revealed-439db1568180"><strong>Read More</strong></a></p>