Differences

This shows you the differences between two versions of the page.

--- ai_timelines:predictions_of_human-level_ai_timelines:ai_timeline_surveys:ai_timeline_surveys [2023/02/15 21:32]
harlanstewart
+++ ai_timelines:predictions_of_human-level_ai_timelines:ai_timeline_surveys:ai_timeline_surveys [2024/01/09 20:12]
harlanstewart
@@ Line 1: / Line 1: @@
 ====== AI Timeline Surveys ======
+/*
+EDITOR COMMENTS:
-// Published 10 January 2015; last updated 12 February 2023 //
+-Harlan: I think that it's a bit inaccurate to say that  Gruetzemacher's survey had a response rate of 40%, because they only sent emails to those who said in person that they would be interested in taking the survey. I'm removing that number from the table
+-Harlan: This page generally could use some work, much of the discussion after the table was written long ago.
+-Harlan: Zach pointed out that we should add something about this survey as well https://www.generationlab.org/expertsai
+*/
+// Published 10 January 2015; last updated 9 January 2024 //
-<HTML>
-<p><em>This page is incomplete, under active work and may be updated soon.</em></p>
-</HTML>
+// This page is incomplete, under active work and may be updated soon. //
-<HTML>
-<p>We know of 18 surveys on the predicted timing of human-level AI. If we collapse a few slightly different meanings of ‘human-level AI’, then:</p>
-</HTML>
+We know of 20 surveys on the predicted timing of human-level AI. If we collapse a few slightly different meanings of ‘human-level AI’, then:
-<HTML>
+  * Median estimates made before 2016 for when there will be a 10% chance of human-level AI are all in the 2020s, median estimates made after 2016 for when there will be a 10% chance of human-level AI are in the 2030s, except for  one with a median of 2046.
-<ul>
+  * Median estimates for when there will be a 50% chance of human-level AI ranged between 2035 and 2074 (from twelve surveys).
-<li><div class="li">Median estimates for when there will be a 10% chance of human-level AI are all in the 2020s (from seven surveys), except for the <span style="font-weight: 400;"><a href="/doku.php?id=ai_timelines:predictions_of_human-level_ai_timelines:ai_timeline_surveys:2016_expert_survey_on_progress_in_ai">2016 ESPAI</a>, which found median estimates ranging from 2013 to long after 2066, depending on question framing</span>.</div></li>
+  * Of three surveys in recent decades asking for predictions but not probabilities, two produced median estimates of when human-level AI will arrive in the 2050s, and one in 2085.
-<li><div class="li">Median estimates for when there will be a 50% chance of human-level AI range between 2035 and 2050 (from seven surveys), except for the <span style="font-weight: 400;"><a href="/doku.php?id=ai_timelines:predictions_of_human-level_ai_timelines:ai_timeline_surveys:2016_expert_survey_on_progress_in_ai">2016 ESPAI</a>, which found median estimates ranging from 2056 to at least 2106, depending on question framing.</span></div></li>
-<li><div class="li">Of three surveys in recent decades asking for predictions but not probabilities, two produced median estimates of when human-level AI will arrive in the 2050s, and one in 2085.</div></li>
-</ul>
-</HTML>
+Participants appear to mostly be experts in AI or related areas, but with a large contingent of others. Several groups of survey participants seem likely to over-represent people who are especially optimistic about human-level AI being achieved soon.
-<HTML>
-<p>Participants appear to mostly be experts in AI or related areas, but with a large contingent of others. Several groups of survey participants seem likely over-represent people who are especially optimistic about human-level AI being achieved soon.</p>
-</HTML>
@@ Line 88: / Line 83: @@
 <a href="https://wiki.aiimpacts.org/doku.php?id=ai_timelines:predictions_of_human-level_ai_timelines:ai_timeline_surveys:2022_expert_survey_on_progress_in_ai" title="2022 Expert Survey on Progress in AI">ESPAI 2022</a> (2022)
                 </div></li>
+<li><div class="li">
+<a href="https://wiki.aiimpacts.org/ai_timelines/predictions_of_human-level_ai_timelines/ai_timeline_surveys/2023_expert_survey_on_progress_in_ai" title="2023 Expert Survey on Progress in AI">ESPAI 2023</a> (2023)
+                </div></li>
 </ul>
 </HTML>
@@ Line 98: / Line 96: @@
 === Results summary ===
@@ Line 112: / Line 108: @@
 <td style="width: 65pt;" width="65">Survey</td>
 <td style="width: 65pt;" width="65">#</td>
-<td class="xl64" style="width: 65pt;" width="65"> 10%</td>
+<td class="xl64" style="width: 65pt;" width="65"> 10% chance</td>
-<td class="xl64" style="width: 65pt; font-size: inherit;" width="65"> 50%</td>
+<td class="xl64" style="width: 65pt; font-size: inherit;" width="65"> 50% chance</td>
-<td class="xl64" style="width: 65pt; font-size: inherit;" width="65"> 90%</td>
+<td class="xl64" style="width: 65pt; font-size: inherit;" width="65"> 90% chance</td>
 <td class="xl64" style="width: 65pt; font-size: inherit;" width="65"> Other key ‘Predictions’</td>
 <td class="xl64" style="width: 65pt;" width="65">Participants</td>
@@ Line 121: / Line 117: @@
 </tr>
 <tr style="height: 21.0pt;">
-<td align="right" height="21" style="height: 21.0pt;">1972</td>
+<td height="21" style="height: 21.0pt;">1972</td>
 <td class="xl65" style="font-size: inherit;">
 <a href="http://aiimpacts.wpengine.com/michie-survey/" title="Michie Survey"> Michie</a>
 </td>
-<td align="right">67</td>
+<td>67</td>
 <td class="xl64" style="font-size: inherit;"> </td>
 <td class="xl64" style="font-size: inherit;"> </td>
@@ Line 137: / Line 133: @@
 </tr>
 <tr style="height: 21.0pt;">
-<td align="right" height="21" style="height: 21.0pt;">2005</td>
+<td height="21" style="height: 21.0pt;">2005</td>
 <td class="xl65" style="font-size: inherit;">
 <a href="http://aiimpacts.wpengine.com/bainbridge-survey/" title="Bainbridge survey"> Bainbridge</a>
 </td>
-<td align="right">26</td>
+<td>26</td>
 <td class="xl64" style="font-size: inherit;"> </td>
 <td class="xl64" style="font-size: inherit;"> </td>
@@ Line 153: / Line 149: @@
 </tr>
 <tr style="height: 21.0pt;">
-<td align="right" height="21" style="height: 21.0pt;">2006</td>
+<td height="21" style="height: 21.0pt;">2006</td>
 <td class="xl65" style="font-size: inherit;">
 <a href="http://aiimpacts.wpengine.com/ai50-survey/" title="AI@50 Survey"> AI@50</a>
@@ Line 169: / Line 165: @@
 </tr>
 <tr style="height: 21.0pt;">
-<td align="right" height="21" style="height: 21.0pt;">2007</td>
+<td height="21" style="height: 21.0pt;">2007</td>
 <td class="xl65" style="font-size: inherit;">
 <a href="http://aiimpacts.wpengine.com/klein-agi-survey/" title="Klein AGI Survey"> Klein</a>
 </td>
-<td align="right">888</td>
+<td>888</td>
 <td class="xl64" style="font-size: inherit;"> </td>
 <td class="xl64" style="font-size: inherit;"> </td>
@@ Line 186: / Line 182: @@
 <tr style="height: 21.0pt;">
-<td align="right" height="21" style="height: 21.0pt;">2009</td>
+<td height="21" style="height: 21.0pt;">2009</td>
 <td class="xl65" style="font-size: inherit;">
 <a href="http://aiimpacts.wpengine.com/agi-09-survey/" title="AGI-09 Survey"> AGI-09</a>
@@ Line 203: / Line 199: @@
 <tr style="height: 21.0pt;">
-<td align="right" height="21" style="height: 21.0pt;">2011</td>
+<td height="21" style="height: 21.0pt;">2011</td>
 <td class="xl65" style="font-size: inherit;">
 <a href="https://wiki.aiimpacts.org/doku.php?id=ai_timelines:predictions_of_human-level_ai_timelines:ai_timeline_surveys:agi-11_survey" title="AGI-11 survey"> AGI-11 </a>
 </td>
-<td align="right">60</td>
+<td>60</td>
-<td align="right" class="xl64" style="font-size: inherit;"></td>
+<td class="xl64" style="font-size: inherit;"></td>
 <td class="xl64" style="font-size: inherit;"> </td>
 <td class="xl64" style="font-size: inherit;"> </td>
 <td class="xl64" style="font-size: inherit;"> </td>
 <td>AGI-11</td>
-<td align="right" class="xl66"></td>
+<td class="xl66"></td>
 <td>
 <a href="https://web.archive.org/web/20221129004757/https://hplusmagazine.com/2011/09/16/how-long-till-agi-views-of-agi-11-conference-participants/">link</a>
@@ Line 221: / Line 217: @@
 <tr style="height: 21.0pt;">
-<td align="right" height="21" style="height: 21.0pt;">2011</td>
+<td height="21" style="height: 21.0pt;">2011</td>
 <td class="xl65" style="font-size: inherit;">
 <a href="http://aiimpacts.wpengine.com/fhi-ai-timelines-survey/" title="FHI Winter Intelligence Survey"> FHI Winter Intelligence</a>
 </td>
-<td align="right">35</td>
+<td>35</td>
 <td class="xl64" style="font-size: inherit;"> 2028</td>
-<td align="right" class="xl64" style="font-size: inherit;">2050</td>
+<td class="xl64" style="font-size: inherit;">2050</td>
 <td class="xl64" style="font-size: inherit;"> 2150</td>
 <td class="xl64" style="font-size: inherit;"> </td>
 <td class="xl64">AGI impacts conf; 44% related technical</td>
-<td align="right" class="xl66">41%</td>
+<td class="xl66">41%</td>
 <td>
 <a href="https://www.fhi.ox.ac.uk/wp-content/uploads/2011-1.pdf">link</a>
@@ Line 237: / Line 233: @@
 </tr>
 <tr style="height: 21.0pt;">
-<td height="21" style="height: 21.0pt;" align= "right">2011-2012</td>
+<td height="21" style="height: 21.0pt;">2011-2012</td>
 <td class="xl65" style="font-size: inherit;">
 <a href="http://aiimpacts.wpengine.com/kruel-ai-survey/" title="Kruel AI Interviews"> Kruel interviews</a>
 </td>
-<td align="right">37</td>
+<td >37</td>
 <td class="xl64" style="font-size: inherit;"> 2025</td>
 <td class="xl64" style="font-size: inherit;"> 2035</td>
@@ Line 253: / Line 249: @@
 </tr>
 <tr style="height: 21.0pt;">
-<td align="right" height="21" style="height: 21.0pt;">2012</td>
+<td height="21" style="height: 21.0pt;">2012</td>
 <td class="xl65" style="font-size: inherit;">
 <a href="http://aiimpacts.wpengine.com/muller-and-bostrom-ai-progress-poll/" title="Müller and Bostrom AI Progress Poll"> FHI: AGI-12</a>
 </td>
-<td align="right">72</td>
+<td >72</td>
 <td class="xl64" style="font-size: inherit;"> 2022</td>
 <td class="xl64" style="font-size: inherit;"> 2040</td>
@@ Line 263: / Line 259: @@
 <td class="xl64" style="font-size: inherit;"> </td>
 <td class="xl64">AGI &amp; AGI impacts conf; AGI, technical work</td>
-<td align="right" class="xl66">65%</td>
+<td class="xl66">65%</td>
 <td>
 <a href="http://www.nickbostrom.com/papers/survey.pdf">link</a>
@@ Line 269: / Line 265: @@
 </tr>
 <tr style="height: 21.0pt;">
-<td align="right" height="21" style="height: 21.0pt;">2012</td>
+<td height="21" style="height: 21.0pt;">2012</td>
 <td class="xl65" style="font-size: inherit;">
 <a href="http://aiimpacts.wpengine.com/muller-and-bostrom-ai-progress-poll/" title="Müller and Bostrom AI Progress Poll"> FHI:PT-AI</a>
 </td>
-<td align="right">43</td>
+<td>43</td>
 <td class="xl64" style="font-size: inherit;"> 2023</td>
 <td class="xl64" style="font-size: inherit;"> 2048</td>
@@ Line 279: / Line 275: @@
 <td class="xl64" style="font-size: inherit;"> </td>
 <td class="xl64">Philosophy &amp; theory of AI conf; not technical AI</td>
-<td align="right" class="xl66">49%</td>
+<td class="xl66">49%</td>
 <td>
 <a href="http://www.nickbostrom.com/papers/survey.pdf">link</a>
@@ Line 285: / Line 281: @@
 </tr>
 <tr style="height: 21.0pt;">
-<td align = "right" height="21" style="height: 21.0pt;">2012-?</td>
+<td height="21" style="height: 21.0pt;">2012-?</td>
 <td class="xl65" style="font-size: inherit;">
 <a href="http://aiimpacts.wpengine.com/hanson-ai-expert-survey/" title="Hanson AI Expert Survey"> Hanson</a>
@@ Line 301: / Line 297: @@
 </tr>
 <tr style="height: 21.0pt;">
-<td align="right" height="21" style="height: 21.0pt;">2013</td>
+<td height="21" style="height: 21.0pt;">2013</td>
 <td class="xl65" style="font-size: inherit;">
 <a href="http://aiimpacts.wpengine.com/muller-and-bostrom-ai-progress-poll/" title="Müller and Bostrom AI Progress Poll"> FHI: TOP100</a>
 </td>
-<td align="right">29</td>
+<td >29</td>
-<td align="right" class="xl64" style="font-size: inherit;">2022</td>
+<td class="xl64" style="font-size: inherit;">2022</td>
 <td class="xl64" style="font-size: inherit;"> 2040</td>
 <td class="xl64" style="font-size: inherit;"> 2075</td>
 <td class="xl64" style="font-size: inherit;"> </td>
 <td>Top AI</td>
-<td align="right" class="xl66">29%</td>
+<td class="xl66">29%</td>
 <td>
 <a href="http://www.nickbostrom.com/papers/survey.pdf">link</a>
@@ Line 317: / Line 313: @@
 </tr>
 <tr style="height: 21.0pt;">
-<td align="right" height="21" style="height: 21.0pt;">2013</td>
+<td height="21" style="height: 21.0pt;">2013</td>
 <td class="xl65" style="font-size: inherit;">
 <a href="http://aiimpacts.wpengine.com/muller-and-bostrom-ai-progress-poll/" title="Müller and Bostrom AI Progress Poll"> FHI:EETN</a>
 </td>
-<td align="right">26</td>
+<td >26</td>
 <td class="xl64" style="font-size: inherit;"> 2020</td>
 <td class="xl64" style="font-size: inherit;"> 2050</td>
@@ Line 327: / Line 323: @@
 <td class="xl64" style="font-size: inherit;"> </td>
 <td class="xl64">Greek assoc. for AI; AI</td>
-<td align="right" class="xl66">10%</td>
+<td class="xl66">10%</td>
 <td>
 <p><a href="http://www.nickbostrom.com/papers/survey.pdf">link</a></p>
@@ Line 333: / Line 329: @@
 </tr>
 <tr style="height: 21.0pt;">
-<td align="right" height="21" style="height: 21.0pt;">2015</td>
+<td height="21" style="height: 21.0pt;">2015</td>
 <td class="xl65" style="font-size: inherit;">
 <a href="https://wiki.aiimpacts.org/doku.php?id=ai_timelines:predictions_of_human-level_ai_timelines:ai_timeline_surveys:2016_expert_survey_on_progress_in_ai" title="2016 Expert Survey on Progress in AI"> ESPAI 2016</a>
 </td>
-<td align="right">352</td>
+<td >352</td>
-<td align="right" class="xl64" style="font-size: inherit;">2025</td>
+<td  class="xl64" style="font-size: inherit;">2025</td>
 <td class="xl64" style="font-size: inherit;"> 2061</td>
 <td class="xl64" style="font-size: inherit;"> </td>
 <td class="xl64" style="font-size: inherit;"> </td>
 <td>NIPS and ICML 2015</td>
-<td align="right" class="xl66">21%</td>
+<td class="xl66">21%</td>
 <td>
 <a href="https://wiki.aiimpacts.org/doku.php?id=ai_timelines:predictions_of_human-level_ai_timelines:ai_timeline_surveys:2016_expert_survey_on_progress_in_ai">link</a>
@@ Line 349: / Line 345: @@
 </tr>
 <tr style="height: 21.0pt;">
-<td align="right" height="21" style="height: 21.0pt;">2016</td>
+<td height="21" style="height: 21.0pt;">2016</td>
 <td class="xl65" style="font-size: inherit;">
 <a href="https://wiki.aiimpacts.org/doku.php?id=ai_timelines:predictions_of_human-level_ai_timelines:ai_timeline_surveys:etzioni_2016_survey" title="Etzioni 2016 survey"> Etzioni</a>
 </td>
-<td align="right">80</td>
+<td>80</td>
-<td align="right" class="xl64" style="font-size: inherit;"></td>
+<td class="xl64" style="font-size: inherit;"></td>
 <td class="xl64" style="font-size: inherit;"> </td>
 <td class="xl64" style="font-size: inherit;"> </td>
 <td class="xl64" style="font-size: inherit;"> </td>
 <td></td>
-<td align="right" class="xl66">41%</td>
+<td class="xl66">41%</td>
 <td>
 <a href="https://wiki.aiimpacts.org/doku.php?id=ai_timelines:predictions_of_human-level_ai_timelines:ai_timeline_surveys:etzioni_2016_survey">link</a>
@@ Line 365: / Line 361: @@
 </tr>
 <tr style="height: 21.0pt;">
-<td align="right" height="21" style="height: 21.0pt;">2017</td>
+<td height="21" style="height: 21.0pt;">2017</td>
 <td class="xl65" style="font-size: inherit;">
 <a href="https://wiki.aiimpacts.org/doku.php?id=ai_timelines:predictions_of_human-level_ai_timelines:ai_timeline_surveys:walsh_2017_survey" title="Walsh 2017 survey"> Walsh</a>
 </td>
-<td align="right">849</td>
+<td >849</td>
 <td class="xl64" style="font-size: inherit;">AI experts: 2035<BR>Robotics experts: 2033<BR>Non-experts: 2026</td>
 <td class="xl64" style="font-size: inherit;">AI experts: 2061<BR>Robotics experts: 2065<BR>Non-experts: 2039</td>
@@ Line 375: / Line 371: @@
 <td class="xl64" style="font-size: inherit;"> </td>
 <td>200 AI Experts<BR>101 Robotics experts<BR>548 Non-experts</td>
-<td align="right" class="xl66"></td>
+<td class="xl66"></td>
 <td>
 <a href="https://wiki.aiimpacts.org/doku.php?id=ai_timelines:predictions_of_human-level_ai_timelines:ai_timeline_surveys:walsh_2017_survey">link</a>
@@ Line 381: / Line 377: @@
 </tr>
 <tr style="height: 21.0pt;">
-<td align="right" height="21" style="height: 21.0pt;">2018</td>
+<td height="21" style="height: 21.0pt;">2018</td>
 <td class="xl65" style="font-size: inherit;">
 <a href="https://arxiv.org/pdf/1901.08579.pdf" title="Forecasting Transformative AI: An Expert Survey"> Gruetzemacher</a>
 </td>
-<td align="right">164</td>
+<td>164</td>
-<td align="right" class="xl64" style="font-size: inherit;"></td>
+<td class="xl64" style="font-size: inherit;">2046</td>
-<td class="xl64" style="font-size: inherit;"> </td>
+<td class="xl64" style="font-size: inherit;">2074</td>
-<td class="xl64" style="font-size: inherit;"> </td>
+<td class="xl64" style="font-size: inherit;">2117</td>
 <td class="xl64" style="font-size: inherit;"> </td>
 <td>ICML, IJCAI, and HLAI 2018</td>
-<td align="right" class="xl66">40%</td>
+<td class="xl66"></td>
 <td>
 <a href="https://wiki.aiimpacts.org/doku.php?id=ai_timelines:predictions_of_human-level_ai_timelines:ai_timeline_surveys:2016_expert_survey_on_progress_in_ai">link</a>
@@ Line 397: / Line 393: @@
 </tr>
 <tr style="height: 21.0pt;">
-<td align="right" height="21" style="height: 21.0pt;">2019</td>
+<td height="21" style="height: 21.0pt;">2019</td>
 <td class="xl65" style="font-size: inherit;">
 <a href="https://arxiv.org/abs/2206.04132" title="Forecasting AI Progress: Evidence from a Survey of Machine Learning Researchers"> GovAI</a>
 </td>
-<td align="right">296</td>
+<td>296</td>
-<td align="right" class="xl64" style="font-size: inherit;"></td>
+<class="xl64" style="font-size: inherit;"></td>
-<td class="xl64" style="font-size: inherit;"> </td>
+<td class="xl64" style="font-size: inherit;">2039</td>
-<td class="xl64" style="font-size: inherit;"> </td>
+<td class="xl64" style="font-size: inherit;">2059</td>
-<td class="xl64" style="font-size: inherit;"> </td>
+<td class="xl64" style="font-size: inherit;">2119</td>
-<td>NIPS and ICML 2018</td>
+<td></td>
-<td align="right" class="xl66"></td>
+<td class="xl66">NIPS and ICML 2018</td>
+<td>20%</td>
 <td>
 <a href="https://arxiv.org/abs/2206.04132">link</a>
@@ Line 413: / Line 410: @@
 </tr>
 <tr style="height: 21.0pt;">
-<td align="right" height="21" style="height: 21.0pt;">2022</td>
+<td height="21" style="height: 21.0pt;">2022</td>
 <td class="xl65" style="font-size: inherit;">
 <a href="https://wiki.aiimpacts.org/doku.php?id=ai_timelines:predictions_of_human-level_ai_timelines:ai_timeline_surveys:2022_expert_survey_on_progress_in_ai" title="2022 Expert Survey on Progress in AI"> ESPAI 2022</a>
 </td>
-<td align="right">738</td>
+<td>461</td>
-<td align="right" class="xl64" style="font-size: inherit;"></td>
+<td class="xl64" style="font-size: inherit;">2029</td>
-<td class="xl64" style="font-size: inherit;"> 2059</td>
+<td class="xl64" style="font-size: inherit;">2060</td>
-<td class="xl64" style="font-size: inherit;"> </td>
+<td class="xl64" style="font-size: inherit;">2288</td>
 <td class="xl64" style="font-size: inherit;"> </td>
 <td>NIPS and ICML 2021</td>
-<td align="right" class="xl66">17%</td>
+<td class="xl66">17%</td>
 <td>
 <a href="https://wiki.aiimpacts.org/doku.php?id=ai_timelines:predictions_of_human-level_ai_timelines:ai_timeline_surveys:2022_expert_survey_on_progress_in_ai">link</a>
 </td>
 </tr>
+<tr style="height: 21.0pt;">
+<td height="21" style="height: 21.0pt;">2023</td>
+<td class="xl65" style="font-size: inherit;">
+<a href="https://wiki.aiimpacts.org/ai_timelines/predictions_of_human-level_ai_timelines/ai_timeline_surveys/2023_expert_survey_on_progress_in_ai" title="2023 Expert Survey on Progress in AI"> ESPAI 2023</a>
+</td>
+<td>1714</td>
+<td class="xl64" style="font-size: inherit;">2027</td>
+<td class="xl64" style="font-size: inherit;">2047</td>
+<td class="xl64" style="font-size: inherit;">2186</td>
+<td class="xl64" style="font-size: inherit;"> </td>
+<td>NeurIPS, ICML, ICLR, AAAI, JMLR, and IJCAI 2022</td>
+<td class="xl66">15%</td>
+<td>
+<a href="https://wiki.aiimpacts.org/ai_timelines/predictions_of_human-level_ai_timelines/ai_timeline_surveys/2023_expert_survey_on_progress_in_ai">link</a>
+</td>
+</tr>
 </tbody>
 </table>

AI Impacts Wiki

User Tools

Site Tools

Differences

Page Tools