Embed Notice

HTML Code

<blockquote style="position: relative; padding-left: 55px;"><section><a href="https://fedi.simonwillison.net/users/simon/statuses/113624603777495615">Simon Willison (simon@fedi.simonwillison.net)'s status on Tuesday, 10-Dec-2024 04:53:10 JST</a><a href="https://fedi.simonwillison.net/@simon" title="simon@fedi.simonwillison.net"><img src="https://gnusocial.jp/avatar/18625-48-20221108165340.webp" width="48" height="48" alt="Simon Willison" style="position: absolute; left: 0; top: 0;">Simon Willison</a><div><a href="https://fedi.simonwillison.net/@simon/113623535989351238" rel="in-reply-to">in reply to</a></div></section><article><p>This consistent trend of models getting smaller, faster and more capable on the same hardware is one of the reasons I'm not particularly concerned by the ongoing discourse about models hitting a plateau<br><a href="https://simonwillison.net/2024/Dec/9/llama-33-70b/#is-performance-about-to-plateau-" rel="nofollow noreferrer">https://simonwillison.net/2024/Dec/9/llama-33-70b/#is-performance-about-to-plateau-</a></p></article><footer><a rel="bookmark" href="https://gnusocial.jp/conversation/4154498#notice-8119978">In conversation</a><time datetime="2024-12-10T04:53:10+09:00" title="Tuesday, 10-Dec-2024 04:53:10 JST">about 4 months ago</time> <span>from <span><a href="https://fedi.simonwillison.net/@simon/113624603777495615" rel="external" title="Sent from fedi.simonwillison.net via ActivityPub">fedi.simonwillison.net</a></span></span><a href="https://fedi.simonwillison.net/@simon/113624603777495615">permalink</a><h4>Attachments</h4><ol><li><label><a rel="external" href="https://gnusocial.jp/attachment/3664640">Is performance about to plateau? #I’ve been mostly unconvinced by the ongoing discourse around LLMs hitting a plateau. The areas I’m personally most excited about are multi-modality (images, audio and video as input) and model efficiency. Both of those have had enormous leaps forward in the past year.I don’t particularly care about “AGI”. I want models that can do useful things that I tell them to, quickly and inexpensively—and that’s exactly what I’ve been getting more of over the past twelve months.Even if progress on these tools entirely stopped right now, the amount I could get done with just the models I’ve downloaded and stashed on a USB drive would keep me busy and productive for years.</a></label><br><a href="https://cdn.masto.host/fedisimonwillisonnet/media_attachments/files/113/624/603/584/330/272/original/5269aa79cc9eec0d.png" rel="external">https://cdn.masto.host/fedisimonwillisonnet/media_attachments/files/113/624/603/584/330/272/original/5269aa79cc9eec0d.png</a></li><li><article><header><div>Domain not in remote thumbnail source whitelist: static.simonwillison.net</div><h5><a href="https://simonwillison.net/2024/Dec/9/llama-33-70b/#is-performance-about-to-plateau-">I can now run a GPT-4 class model on my laptop</a></h5><div> from <span>@simonw</span></div></header><div>Meta’s new Llama 3.3 70B is a genuinely GPT-4 class Large Language Model that runs on my laptop. Just 20 months ago I was amazed to see something that felt …</div><footer></footer></article></li></ol></footer></blockquote>

Corresponding Notice

Embed this notice
Simon Willison (simon@fedi.simonwillison.net)'s status on Tuesday, 10-Dec-2024 04:53:10 JST Simon Willison
in reply to
This consistent trend of models getting smaller, faster and more capable on the same hardware is one of the reasons I'm not particularly concerned by the ongoing discourse about models hitting a plateau
https://simonwillison.net/2024/Dec/9/llama-33-70b/#is-performance-about-to-plateau-
In conversationabout 4 months ago from fedi.simonwillison.netpermalink
Attachments
1. Is performance about to plateau? # I’ve been mostly unconvinced by the ongoing discourse around LLMs hitting a plateau. The areas I’m personally most excited about are multi-modality (images, audio and video as input) and model efficiency. Both of those have had enormous leaps forward in the past year. I don’t particularly care about “AGI”. I want models that can do useful things that I tell them to, quickly and inexpensively—and that’s exactly what I’ve been getting more of over the past twelve months. Even if progress on these tools entirely stopped right now, the amount I could get done with just the models I’ve downloaded and stashed on a USB drive would keep me busy and productive for years.
  https://cdn.masto.host/fedisimonwillisonnet/media_attachments/files/113/624/603/584/330/272/original/5269aa79cc9eec0d.png
2. Domain not in remote thumbnail source whitelist: static.simonwillison.net
  I can now run a GPT-4 class model on my laptop
  from @simonw
  Meta’s new Llama 3.3 70B is a genuinely GPT-4 class Large Language Model that runs on my laptop. Just 20 months ago I was amazed to see something that felt …

Public

Embed Notice

HTML Code

Corresponding Notice