![\"\"](\"https:\/\/world.edu\/wp-content\/uploads\/2025\/01\/synthetic-data.jpeg\")
<\/p>\n<\/p>\n
Last week the billionaire and owner of X, Elon Musk,\u00a0claimed<\/a>\u00a0the pool of human-generated data that\u2019s used to train artificial intelligence (AI) models such as ChatGPT has run out.<\/p>\n Musk didn\u2019t cite evidence to support this. But other leading tech industry figures have made\u00a0similar claims<\/a>\u00a0in recent months. And\u00a0earlier research<\/a>\u00a0indicated human-generated data would run out within two to eight years.<\/p>\n This is largely because humans can\u2019t create new data such as text, video and images fast enough to keep up with the speedy and enormous demands of AI models. When genuine data does run out, it will present a major problem for both developers and users of AI.<\/p>\n It will force tech companies to depend more heavily on data generated by AI, known as \u201csynthetic data\u201d. And this, in turn, could lead to the AI systems currently used by\u00a0hundreds of millions<\/a>\u00a0of people being less accurate and reliable \u2013 and therefore, useful.<\/p>\n But this isn\u2019t an inevitable outcome. In fact, if used and managed carefully, synthetic data could improve AI models.<\/p>\n Tech companies depend on data \u2013 real or synthetic \u2013 to build, train and refine generative AI models such as ChatGPT. The\u00a0quality of this data<\/a>\u00a0is crucial. Poor data leads to poor outputs, in the same way using low-quality ingredients in cooking can produce low-quality meals.<\/p>\n Real data<\/a>\u00a0refers to text, video and images created by humans. Companies collect it through methods such as surveys, experiments, observations or mining of websites and social media.<\/p>\n Real data is generally considered valuable because it includes true events and captures a wide range of scenarios and contexts. However, it isn\u2019t perfect.<\/p>\n For example, it can contain\u00a0spelling errors and inconsistent or irrelevant content<\/a>. It can also be\u00a0heavily biased<\/a>, which can, for example, lead to generative AI models\u00a0creating images<\/a>\u00a0that show only men or white people in certain jobs.<\/p>\n This kind of data also requires a lot of time and effort to prepare. First, people collect datasets,\u00a0before labelling them<\/a>\u00a0to make them meaningful for an AI model. They will then review and clean this data to resolve any inconsistencies, before computers filter, organise and validate it.<\/p>\n This process can take up to\u00a080% of the total time investment<\/a>\u00a0in the development of an AI system.<\/p>\n But as stated above, real data is also in\u00a0increasingly short supply<\/a>\u00a0because humans can\u2019t produce it quickly enough to feed burgeoning AI demand.<\/p>\n Synthetic data<\/a>\u00a0is artificially created or\u00a0generated by algorithms<\/a>, such as text generated by\u00a0ChatGPT<\/a>\u00a0or an image generated by\u00a0DALL-E<\/a>.<\/p>\n In theory, synthetic data offers a cost-effective and faster solution for training AI models.<\/p>\n It also addresses\u00a0privacy concerns and ethical issues<\/a>, particularly with sensitive personal information like health data.<\/p>\n Importantly, unlike real data it isn\u2019t in short supply. In fact, it\u2019s unlimited. From here on its only Synthetic data.<\/p>\n "Cumulative sum of human knowledge has been exhausted in AI training. That happened, basically, last year.\u201d<\/p>\n – Elonpic.twitter.com\/rdPzCbvdLv<\/a><\/p>\n — Rohan Paul (@rohanpaul_ai) January 9, 2025<\/a><\/p><\/blockquote>\n The problems with real data<\/strong><\/h4>\n
The rise of synthetic data<\/strong><\/h4>\n
\n\n