<?xml version="1.0" encoding="UTF-8"?>
<feed xmlns="http://www.w3.org/2005/Atom" xmlns:dw="https://www.dreamwidth.org">
  <id>tag:dreamwidth.org,2017-07-05:3235132</id>
  <title>Dataflow matrix machines (by Anhinga anhinga)</title>
  <subtitle>Dataflow matrix machines (by Anhinga anhinga)</subtitle>
  <author>
    <name>Dataflow matrix machines (by Anhinga anhinga)</name>
  </author>
  <link rel="alternate" type="text/html" href="https://dmm.dreamwidth.org/"/>
  <link rel="self" type="text/xml" href="https://dmm.dreamwidth.org/data/atom"/>
  <updated>2022-08-25T13:43:11Z</updated>
  <dw:journal username="dmm" type="personal"/>
  <entry>
    <id>tag:dreamwidth.org,2017-07-05:3235132:63041</id>
    <link rel="alternate" type="text/html" href="https://dmm.dreamwidth.org/63041.html"/>
    <link rel="self" type="text/xml" href="https://dmm.dreamwidth.org/data/atom/?itemid=63041"/>
    <title>OpenAI posted its approach to alignment research</title>
    <published>2022-08-25T13:43:11Z</published>
    <updated>2022-08-25T13:43:11Z</updated>
    <category term="openai codex"/>
    <category term="understanding internals of ai"/>
    <category term="technological singularity"/>
    <category term="ai safety"/>
    <category term="artificial intelligence"/>
    <dw:security>public</dw:security>
    <dw:reply-count>3</dw:reply-count>
    <content type="html">&lt;a href="https://openai.com/blog/our-approach-to-alignment-research/"&gt;openai.com/blog/our-approach-to-alignment-research/&lt;/a&gt;&lt;br /&gt;&lt;br /&gt;&amp;quot;Our approach to aligning AGI is empirical and iterative. We are  improving our AI systems&amp;rsquo; ability to learn from human feedback and to  assist humans at evaluating AI. Our goal is to build a sufficiently  aligned AI system that can help us solve all other alignment&amp;nbsp;problems.&amp;quot;&lt;br /&gt;&lt;br /&gt;&lt;br /&gt;&lt;br /&gt;&lt;img src="https://www.dreamwidth.org/tools/commentcount?user=dmm&amp;ditemid=63041" width="30" height="12" alt="comment count unavailable" style="vertical-align: middle;"/&gt; comments</content>
  </entry>
  <entry>
    <id>tag:dreamwidth.org,2017-07-05:3235132:62501</id>
    <link rel="alternate" type="text/html" href="https://dmm.dreamwidth.org/62501.html"/>
    <link rel="self" type="text/xml" href="https://dmm.dreamwidth.org/data/atom/?itemid=62501"/>
    <title>Open source code generator (an alternative to OpenAI Codex)</title>
    <published>2022-08-09T03:13:17Z</published>
    <updated>2022-08-09T03:13:17Z</updated>
    <category term="program synthesis"/>
    <category term="artificial intelligence"/>
    <category term="transformers"/>
    <category term="github copilot"/>
    <category term="openai codex"/>
    <category term="understanding internals of ai"/>
    <category term="machine learning"/>
    <dw:security>public</dw:security>
    <dw:reply-count>6</dw:reply-count>
    <content type="html">&lt;a href="https://github.com/salesforce/CodeGen"&gt;github.com/salesforce/CodeGen&lt;/a&gt;&lt;br /&gt;&lt;br /&gt;One can also run one of these models via HuggingFace; it is based on &amp;quot;A Conversational Paradigm for Program Synthesis&amp;quot; paper, &lt;a href="https://arxiv.org/abs/2203.13474"&gt;arxiv.org/abs/2203.13474&lt;/a&gt;&lt;br /&gt;&lt;br /&gt;Someone has even created a fake GitHub Copilot based on that (useful for those who prefer VSCode): &lt;a href="https://github.com/moyix/fauxpilot"&gt;github.com/moyix/fauxpilot&lt;/a&gt;&lt;br /&gt;&lt;br /&gt;&lt;br /&gt;&lt;br /&gt;&lt;img src="https://www.dreamwidth.org/tools/commentcount?user=dmm&amp;ditemid=62501" width="30" height="12" alt="comment count unavailable" style="vertical-align: middle;"/&gt; comments</content>
  </entry>
  <entry>
    <id>tag:dreamwidth.org,2017-07-05:3235132:47121</id>
    <link rel="alternate" type="text/html" href="https://dmm.dreamwidth.org/47121.html"/>
    <link rel="self" type="text/xml" href="https://dmm.dreamwidth.org/data/atom/?itemid=47121"/>
    <title>OpenAI Codex (next generation) - it looks like we are finally "there"</title>
    <published>2021-08-12T15:35:14Z</published>
    <updated>2021-08-12T15:37:07Z</updated>
    <category term="openai codex"/>
    <category term="program synthesis"/>
    <category term="transformers"/>
    <dw:security>public</dw:security>
    <dw:reply-count>26</dw:reply-count>
    <content type="html">There was a live demo of the next generation of OpenAI&amp;nbsp;Codex code-generating software on August 10.&lt;br /&gt;&lt;br /&gt;My impression from it is that &amp;quot;we have finally arrived&amp;quot; - this is a programming tool which seems to be more useful than an extra entry-level software engineer on the team. This starts to address the key bottleneck of our times: our limited ability to create software.&lt;br /&gt;&lt;br /&gt;This was always my threshold: can we create an AI software which can be &amp;quot;hired instead of a junior software engineer&amp;quot;? That was the main temporal uncertainty for me: how long would it take to reach that level? It looks like this has been accomplished.&lt;br /&gt;&lt;br /&gt;We are rapidly approaching the situation when AI will actively participate in programming AI software, for better or for worse...&lt;br /&gt;&lt;br /&gt;OpenAI&amp;nbsp;Codex is now a part of OpenAI&amp;nbsp;API&amp;nbsp;(which is still a closed beta with a waitlist), and it will be possible to participate in an informal competition today from 10am Pacific time (1pm Eastern) till 1pm Pacific (4pm Eastern) and try it a bit.&lt;br /&gt;&lt;br /&gt;Links are in the comments.&lt;br /&gt;&lt;br /&gt;&lt;img src="https://www.dreamwidth.org/tools/commentcount?user=dmm&amp;ditemid=47121" width="30" height="12" alt="comment count unavailable" style="vertical-align: middle;"/&gt; comments</content>
  </entry>
  <entry>
    <id>tag:dreamwidth.org,2017-07-05:3235132:44860</id>
    <link rel="alternate" type="text/html" href="https://dmm.dreamwidth.org/44860.html"/>
    <link rel="self" type="text/xml" href="https://dmm.dreamwidth.org/data/atom/?itemid=44860"/>
    <title>GitHub Copilot ("we are getting there")</title>
    <published>2021-06-29T16:23:30Z</published>
    <updated>2021-06-29T16:26:46Z</updated>
    <category term="openai codex"/>
    <category term="artificial intelligence"/>
    <category term="transformers"/>
    <category term="github copilot"/>
    <category term="technological singularity"/>
    <category term="program synthesis"/>
    <dw:security>public</dw:security>
    <dw:reply-count>5</dw:reply-count>
    <content type="html">&lt;a href="https://github.blog/2021-06-29-introducing-github-copilot-ai-pair-programmer/"&gt;github.blog/2021-06-29-introducing-github-copilot-ai-pair-programmer/&lt;/a&gt;&lt;br /&gt;&lt;br /&gt;&amp;quot;Today, we are launching a technical preview of &lt;a href="http://copilot.github.com" target="_blank" rel="noopener"&gt;GitHub Copilot&lt;/a&gt;,  a new AI pair programmer that helps you write better code. GitHub  Copilot draws context from the code you&amp;rsquo;re working on, suggesting whole  lines or entire functions. It helps you quickly discover alternative  ways to solve problems, write tests, and explore new APIs without having  to tediously tailor a search for answers on the internet. As you type,  it adapts to the way you write code&amp;mdash;to help you complete your work  faster. &lt;p&gt;Developed in collaboration with OpenAI, GitHub Copilot is powered by  OpenAI Codex, a new AI system created by OpenAI. OpenAI Codex has broad  knowledge of how people use code and is significantly more capable than  GPT-3 in code generation, in part, because it was trained on a data set  that includes a much larger concentration of public source code. GitHub  Copilot works with a broad set of frameworks and languages, but this  technical preview works especially well for Python, JavaScript,  TypeScript, Ruby and Go.&amp;quot;&lt;br /&gt;&lt;br /&gt;If you are using Visual Studio Code often, it might make sense to try to sign-up for the technical preview phase...&lt;/p&gt;&lt;br /&gt;&lt;br /&gt;&lt;img src="https://www.dreamwidth.org/tools/commentcount?user=dmm&amp;ditemid=44860" width="30" height="12" alt="comment count unavailable" style="vertical-align: middle;"/&gt; comments</content>
  </entry>
</feed>
