/ricerca/ansaen/search.shtml?any=
Show less

Se hai scelto di non accettare i cookie di profilazione e tracciamento, puoi aderire all’abbonamento "Consentless" a un costo molto accessibile, oppure scegliere un altro abbonamento per accedere ad ANSA.it.

Ti invitiamo a leggere le Condizioni Generali di Servizio, la Cookie Policy e l'Informativa Privacy.

Puoi leggere tutti i titoli di ANSA.it
e 10 contenuti ogni 30 giorni
a €16,99/anno

  • Servizio equivalente a quello accessibile prestando il consenso ai cookie di profilazione pubblicitaria e tracciamento
  • Durata annuale (senza rinnovo automatico)
  • Un pop-up ti avvertirà che hai raggiunto i contenuti consentiti in 30 giorni (potrai continuare a vedere tutti i titoli del sito, ma per aprire altri contenuti dovrai attendere il successivo periodo di 30 giorni)
  • Pubblicità presente ma non profilata o gestibile mediante il pannello delle preferenze
  • Iscrizione alle Newsletter tematiche curate dalle redazioni ANSA.


Per accedere senza limiti a tutti i contenuti di ANSA.it

Scegli il piano di abbonamento più adatto alle tue esigenze.

Stanford University outlines latest AI trends

Stanford University outlines latest AI trends

Generative AI investment skyrockets, AI Index Report says

ROME, 08 May 2024, 14:27

ANSA English Desk

ANSACheck

- ALL RIGHTS RESERVED

The Institute for Human-Centered Artificial Intelligence (HAI) of Stanford University has published the seventh edition of the AI Index Report, a study analysing the trends of artificial intelligence (AI), from technological progress and governance to the impact on the economy.
    Here are the main takeaways from the study released by HAI.
    AI beats humans on some tasks, but not on all.
    AI has surpassed human performance on several benchmarks, including some in image classification, visual reasoning, and English understanding, the institute said in a summary of the report published on its website.
    However, it trails behind on more complex tasks including competition-level mathematics, visual commonsense reasoning and planning.
    The industry continues to dominate frontier AI research, HAI found.
    In 2023, the industry produced 51 worthy machine learning models, while academia contributed only 15, the institute explained.
    Moreover, in 2023, 21 notable models were produced thanks to industry-academia cooperation in a new high, according to the report.
    Frontier models meanwhile have become much more expensive.
    According to AI Index estimates, the training costs of state-of-the-art AI models have reached unprecedented levels.
    In particular, the institute cited as an example OpenAI's GPT-4 which it said required an estimated 78 million dollars to train, while Goggle's Gemini Ultra cost 191 million dollars.
    The United States leads China, the EU and Great Britain as the leading model source of top AI models, according to the report.
    In 2023, 61 AI models came from US-based institutions, significantly outpacing the European Union's 21 and China's 15.
    The report noted that robust and standardized evaluations for large language model (LLM) responsibility are seriously lacking.
    The research revealed a significant lack of standardization in responsible AI reporting.
    The main developers, including OpenAI, Google and Anthropic, mainly test their models against different responsible AI benchmarks.
    Such a practice complicates efforts to compare in a systematic way the risks and limitations of top AI models, HAI explained.
    Generative AI investments have skyrocketed.
    In spite of a decline in AI private investments last year, the report found that funding for generative AI surged, almost octupling from 2022 to reach 25.2 billion dollars.
    The report also found that AI makes workers more productive and leads to higher quality work.
    In 2023, different studies evaluated AI's impact on labour, suggesting that AI enables workers to finish tasks more quickly and to improve the quality of their results.
    These studies also demonstrated AI's potential to bridge the gap between low and high-skilled workers, according to HAI.
    However, other studies warned that the use of AI without adequate supervision could lead to diminished performance.
    Scientific progress accelerates even further, thanks to AI, the report found.
    In 2022, AI began to advance scientific discovery.
    The year 2023, however, witnessed the launch of even more relevant science-related AI applications, from AlphaDev, which makes algorithmic sorting more efficient, to GNoME, which aids the process of materials discovery.
    The number of AI regulations in the United States has increased significantly, the report noted.
    The number of regulations related to AI in the US has risen significantly in the past year and over the last five years.
    In 2023 alone, the total number of AI-related regulations rose by 56.3 percent, HAI said.
    People across the world are more aware of AI's potential impact and more nervous about it.
    A survey from the Ipsos research agency highlighted that, over the past year, the proportion of those who think AI will drastically affect their lives over the next three to five years has increased from 60 percent to 66 percent.
    Moreover, 52 percent expressed nervousness toward AI products and services, up 13 percentage points from 2022.
    In the US, Pew data suggested that 52 percent of Americans felt more concerned than excited about AI, up from 38 percent in 2022, HAI concluded.
   

ALL RIGHTS RESERVED © Copyright ANSA

Not to be missed

Share

Or use

ANSA Corporate

If it is news,
it is an ANSA.

We have been collecting, publishing and distributing journalistic information since 1945 with offices in Italy and around the world. Learn more about our services.