LevelsofAGI
ofAGI ontology,whichoffersa morenuancedway to de- October20,2023.
fineprogresstowardAGIbyconsideringgenerality(either
Narrow or General) in tandem with five levels of perfor- Andreas, J., Begusˇ, G., Bronstein, M. M., Diamant, R.,
mance (Emerging, Competent, Expert, Virtuoso, and Su- Delaney, D., Gero, S., Goldwasser, S., Gruber, D. F.,
perhuman). We reflected on how current AI systems and de Haas, S., Malkin, P., Pavlov, N., Payne, R., Petri,
AGIdefinitionsfitintothisframing. Further,wediscussed G., Rus, D., Sharma, P., Tchernov, D., Tønnesen,
the implications of our principles for developing a living, P., Torralba, A., Vogt, D., and Wood, R. J. Toward
ecologicallyvalidAGIbenchmark,andarguethatsuchan understanding the communication in sperm whales.
endeavor, while sure to be challenging, is vital to engage iScience, 25(6):104393, 2022. ISSN 2589-0042.
with. doi: https://doi.org/10.1016/j.isci.2022.104393. URL
https://www.sciencedirect.com/science/article/pii/S
Finally, we considered how our principles and ontology
can reshape discussions around the risks associated with Anil, R., Dai, A. M., Firat, O., and et al. PaLM
AGI.Notably,weobservedthatAGIisnotnecessarilysyn- 2 Technical Report. CoRR, abs/2305.10403,
onymouswithautonomy. WeintroducedLevelsofAuton- 2023. doi: 10.48550/arXiv.2305.10403. URL
omythatareunlocked,butnotdeterminedby,progression https://arxiv.org/abs/2305.10403.
throughtheLevelsofAGI.Weillustratedhowconsidering
AGILeveljointlywithAutonomyLevelcanprovidemore Anthropic. Company: Anthropic, 2023a. URL
nuancedinsightsintorisksassociatedwithAIsystems,un- https://www.anthropic.com/company. Ac-
derscoringthe importanceof investing in human-AIinter- cessedOctober12,2023.
actionresearchintandemwithmodelimprovements.
Anthropic. Anthropic’s Responsible Scal-
We hopeourframeworkwillproveadaptableandscalable
ing Policy, September 2023b. URL
– for instance, how we define and measure progress to- https://www-files.anthropic.com/production/files/res
ward AGI might change with technical advances such as
accessedonOctober20,2023.
improvements in interpretability that provide insight into
models’ inner workings. Additionally, parts of our ontol- Apple.Siri.URLhttps://www.apple.com/siri/.
ogy such as Human-AI Interaction paradigms and associ- accessedonOctober20,2023.
atedrisksmightevolveassocietyitselfadaptstoadvances
inAI. Bellier, L., Llorens, A., Marciano, D., Gunduz, A.,
Schalk, G., Brunner, P., and Knight, R. T. Music can
be reconstructed from human auditory cortex activity
ImpactStatement
usingnonlineardecodingmodels. PLOSBiology,21(8):
This position paper introduces a novel ontology that sup- 1–27,082023.doi:10.1371/journal.pbio.3002176.URL
portsdiscussingprogresstowardAGIinanuancedmanner, https://doi.org/10.1371/journal.pbio.3002176.
withtheaimofsupportingclearcommunicationamongre-
Bengio, Y., Hinton, G., Yao, A., Song, D., Abbeel, P.,
searchers, practitioners, and policymakers about systems’
Harari, Y. N., Zhang, Y.-Q., Xue, L., Shalev-Shwartz,
capabilitiesandassociatedrisks.
S., Hadfield, G., Clune, J., Maharaj, T., Hutter, F.,
Baydin, A. G., McIlraith, S., Gao, Q., Acharya, A.,
Acknowledgements
Krueger, D., Dragan, A., Torr, P., Russell, S., Kah-
neman, D., Brauner, J., and Mindermann, S. Man-
ThankyoutothemembersoftheGoogleDeepMindPAGI
aging AI Risks in an Era of Rapid Progress. CoRR,
team for their support of this effort, and to Martin Wat-
abs/2310.17688,2023.doi:10.48550/arXiv.2310.17688.
tenberg,MichaelTerry,GeoffreyIrving,MurrayShanahan,
URLhttps://arxiv.org/abs/2310.17688.
DileepGeorge,Blaise Agu¨erayArcas,andBenShneider-
manforhelpfuldiscussionsaboutthistopic.
Boden,M.A. GOFAI,pp.89–107. CambridgeUniversity
Press,2014.
References
Brohan, A., Brown, N., Carbajal, J., Chebotar, Y., Chen,
Agu¨erayArcas,B.andNorvig,P. ArtificialGeneralIntel-
X., Choromanski, K., Ding, T., Driess, D., Dubey, A.,
ligence is Already Here. Noema, October 2023. URL
Finn, C., Florence,P., Fu, C., Arenas, M.G., Gopalakr-
https://www.noemamag.com/artificial-general-intelligence-is-already-here/.
ishnan, K., Han, K., Hausman, K., Herzog, A., Hsu, J.,
Ichter, B., Irpan, A., Joshi, N., Julian, R., Kalashnikov,
Amazon. Amazon Alexa. URL D.,Kuang,Y.,Leal,I.,Lee,L.,Lee,T.-W.E.,Levine,S.,
https://alexa.amazon.com/. accessed on Lu,Y.,Michalewski,H.,Mordatch,I.,Pertsch,K.,Rao,
10