The curious case of uncurious creation

ABSTRACTThis paper seeks to answer the question: Can contemporary forms of artificial intelligence be creative? To answer this question, I consider three conditions that are commonly taken to be necessary for creativity. These are novelty, value, and agency. I argue that while contemporary AI models may have a claim to novelty and value, they cannot satisfy the kind of agency condition required for creativity. From this discussion, a new condition for creativity emerges. Creativity requires curiosity, a motivation to pursue epistemic goods. I argue that contemporary AI models do not satisfy this new condition. Because they lack both agency and curiosity, it is a mistake to attribute the same sort of creativity to AI that we prize in humans. Finally, I consider the question of whether these AI models stand to make human creativity in the arts and sciences obsolete, despite not being creative themselves. I argue, optimistically, that this is unlikely.KEYWORDS: Creativityartificial intelligencecuriosityaestheticsagency Disclosure statementNo potential conflict of interest was reported by the author(s).Notes1 I am grateful to Ian Cruise, Marc Lange, Grace-Anna Perry, Nathaniel Sharadin, Keshav Singh, and an anonymous reviewer for their helpful feedback on earlier drafts of this paper.2 For a historical overview of early work in AI, see Crevier (Citation1993) and McCorduck (Citation2004).3 See Bubeck et al. (Citation2023) for some illustrative examples.4 This declaration is the title of a post on Bill Gates’ blog, GatesNotes. See Gates (Citation2023).5 One exception to this is the Zhuangzist account of creativity drawn out by Julianne Chung (Citation2020, Citation2021, Citation2022). On Chung’s interpretation, creativity as conceived by Zhuangzi doesn’t require novelty, but instead requires spontaneity.6 Among those who argue that creativity requires novelty are Beardsley (Citation1965), Boden (Citation1992, Citation2005, Citation2010), Hills and Bird (Citation2019), Gaut (Citation2003, Citation2018), Kronfeldner (Citation2009, Citation2018), Miller (Citation2019), Simonton (Citation1999), and Stokes (Citation2008, Citation2011, Citation2014).7 See, for instance, David Novitz (Citation1999), who argues that creative work must be surprising in the sense that it could not have been easily predicted by the community in which the creative work occurs. Relatedly, Robert Audi (Citation2018) also claims that unpredictability is the crucial sort of novelty for creativity, but argues that it should not be equated with surprisingness.8 Of course, the relevant sort of novelty here might not only be a matter of whether the sentence has been composed before – either historically or psychologically. It might also be a matter of how similar it is to other sentences that have been composed before. This suggests that novelty comes in degrees. For further discussion of this idea, see Brainard (Citationms). I am grateful to Marc Lange for suggesting I address this.9 For some accounts that include value as a necessarily condition of creativity, see Amabile (Citation1996), Boden (Citation1992, Citation2005, Citation2010), Carruthers (Citation2011), Cropley and Cropley (Citation2013), Gaut (Citation2003, Citation2018), Kieran (Citation2014), and Novitz (Citation1999).10 Some philosophers have cast doubt on the claim that creativity is necessarily valuable. Specifically, Hills and Bird (Citation2018; Citation2019) have argued that creativity is not necessarily valuable because there are clear cases of creativity that are worthless as well as clear cases of creativity that are wholly bad. Gaut (Citation2018) has argued that creativity’s value is merely conditional.11 For a helpful overview of epistemic value, see Pritchard and Turri (Citation2018). For a discussion of epistemic value in the context of creativity, see Hawley (Citation2018).12 Monroe Beardsley argues for a similar claim about artistic creativity (Citation1965). On his view, artistic creativity involves unexpected changes and the author’s critical reflection and response to those changes. For this reason, the result of artistic creation can’t be fully anticipated before the process is complete.13 Some accounts of creativity build epistemic value into the definition directly. For instance, Arthur I. Miller defines creativity as ‘the production of new knowledge from already existing knowledge [which] is accomplished by problem solving.’ (Citation2019, 29)14 For an argument that creativity always has the epistemic value of understanding, see Brainard (Citationms).15 For a helpful explanation of why deep learning is a black box, see Blazek (Citation2022).16 See, for example, Brainard (Citationms.), Carruthers (Citation2006), Gaut (Citation2018), Kieran (Citation2014) Paul and Stokes (Citation2018), and Stokes (Citation2008, Citation2011, Citation2014). To my knowledge, the only argument against the claim that creativity requires agency is given by Currie and Turner (Citationforthcoming), who argue that evolutionary processes can be creative.17 I am grateful to Keshav Singh for this example and the point it illustrates.18 Berys Gaut (Citation2018) also argues that creativity requires an element of spontaneity.19 See Bengio, Goodfellow, and Courville (Citation2016) for an explanation of how deep learning models improve iteratively.20 Though the relevant notion of responsibility is stronger than mere causal responsibility, it is not moral responsibility, but rather a more generic form of agential responsibility (see Wolf Citation2015). For an argument that AI cannot achieve moral agency specifically, see Véliz (Citation2021). For discussion of one practical implication of that argument, see Véliz (Citation2023).21 Claire Anscomb (Citation2022) argues that, while AI models that produce images cannot be creative, there is a sense in which they deserve some credit for works of art they contribute to.22 See OpenAI (Citation2023d) for an overview of how human feedback is used to train GPT-4.23 For reasons to be skeptical about categorizing curiosity as a desire, see İnan (Citation2011) and Kvanvig (Citation2012).24 For accounts of curiosity that categorize it as a desire, see Litman and Spielberger (Citation2003), Whitcomb (Citation2010), and Zagzebski (Citation1996).25 The list of philosophers who characterize curiosity as a motivating force includes Baehr (Citation2011), İnan (Citation2011), Kvanvig (Citation2012), Morton (Citation2010), Miščević (Citation2007), Ross (Citation2020), and Whitcomb (Citation2010).26 Christen, Alfano, and Robinson (Citation2014) argue that the inquisitive self is one of the aspects of a person that comprises the disposition of intellectual humility.27 For the original presentation of the distinction between the context of discovery and the context of justification, see Reichenbach (Citation1938). For a historical overview of the way this distinction has usually been understood by philosophers of science, see Schickore (Citation2014).28 See Anfinsen (Citation1973) for an early description of the protein folding problem.29 For an overview of major theories regarding the value and function of scientific explanation, see Salmon (Citation1989).30 See, for instance, Blazek and Lin (Citation2021) and Garcez (Citation2023).