AI makes workers more efficient, but we are still doing not have in regulations, according to brand-new research. The 2024 AI Index Report, released by the Stanford University Human-Centered Artificial Intelligence institute, has actually uncovered the leading eight AI trends for companies, consisting of how the technology still does not best the human brain on every task.
TechRepublic digs into business implications of these takeaways, with insight from report co-authors Robi Rahman and Anka Reuel.
SEE: Top 5 AI Trends to View in 2024
1. Humans still surpass AI on many tasks
According to the research, AI is still not as great as humans at the complex jobs of advanced-level mathematical issue solving, visual commonsense reasoning and preparation (Figure A). To draw this conclusion, designs were compared to human criteria in various service functions, including coding, agent-based behaviour, thinking and support learning.
Figure A
Efficiency of AI models in various tasks relative to human beings. Image: AI Index Report 2024/Stanford University HAI While AI did go beyond human capabilities in image classification, visual thinking and English understanding, the result reveals there is capacity for services to make use of AI for tasks where human personnel would in fact perform better. Numerous businesses are currently concerned about the effects of over-reliance on AI items.
2. Advanced AI designs are getting more pricey
The AI Index reports that OpenAI’s GPT-4 and Google’s Gemini Ultra cost approximately $78 million and $191 million to train in 2023, respectively (Figure B). Information scientist Rahman informed TechRepublic in an e-mail: “At current development rates, frontier AI designs will cost around $5 billion to $10 billion in 2026, at which point extremely few companies will have the ability to manage these training runs.”
Figure B
Training expenses of AI models, 2017 to 2023. Image: AI Index Report 2024/Stanford University HAI/Epoch, 2023 In October 2023, the Wall Street Journal published that Google, Microsoft and other big tech players were struggling to monetize their generative AI products due to the enormous expenses associated with running them. There is a risk that, if the very best innovations end up being so costly that they are exclusively accessible to large corporations, their benefit over SMBs might increase disproportionately. This was flagged by the World Economic Online forum back in 2018. However, Rahman highlighted that a lot of the very best AI models are open source and hence available to organizations of all budget plans, so the innovation should not broaden any gap.
He informed TechRepublic:”Open-source and closed-source AI models are growing at the very same rate. One of the biggest tech business, Meta, is open-sourcing all of their designs, so people who can not pay for to train the largest designs themselves can just download theirs. “3. AI increases productivity and work quality Through examining a number of existing studies, the Stanford researchers concluded that AI makes it possible for workers to finish tasks faster and enhances the
quality of their output. Occupations this was observed for include computer programmers, where 32.8%reported a productivity boost, consultants, assistance representatives (Figure C) and recruiters. Figure C Effect of AI on consumer support agent productivity. Image 
: AI Index Report 2024/Stanford University HAI/Brynjolfsson et al., 2023 When it comes to consultants, using GPT-4 bridged the gap in between low-skilled and high-skilled professionals, with the low-skilled group experiencing more of a performance boost(Figure D ). Other research study has actually also indicated how generative AI in specific could function as an equaliser, as the less knowledgeable, lower knowledgeable workers get more out
of it. Figure D Improvement of work 
performance on low- and high-skilled specialists when utilizing AI. Image: AI Index Report 2024/Stanford University HAI Nevertheless, other research studies did recommend that”utilizing AI without correct oversight can cause lessened performance,”the researchers wrote. For instance, there are prevalent reports that hallucinations are prevalent in large language models that perform legal jobs. Other research has actually found that we may not reach the complete potential of AI-enabled efficiency gains for another decade, as unsatisfactory outputs, complicated guidelines and absence of proficiency continue to hold employees back. 4. AI guidelines in the U.S. are on the increase The AI Index Report discovered that, in 2023, there were 25 AI-related guidelines active in the U.S., while in 2016 there was just one(Figure E). This hasn’t been a consistent slope, however, as the overall number of AI-related guidelines grew by 56.3%from 2022 to 2023 alone. Gradually, these guidelines have actually also moved from being expansive regarding AI development to restrictive, and the most widespread subject they touch on is foreign trade and international finance. Figure E< img src= "https://assets.techrepublic.com/uploads/2024/04/ai-regulations-in-usa.jpg"alt ="Number of AI-related regulations active in the U.S. between 2016 and 2023."width="886"height="480"/ > Number of AI-related regulations active in the U.S. in between 2016 and 
2023. Image: AI Index Report 2024/Stanford University HAI AI-related legislation is also increasing in the EU, with 46, 22 and 32 brand-new guidelines being passed in 2021, 2022 and 2023, respectively. In this region, guidelines tend to take a more expansive approach and usually cover science, technology and interactions. SEE: NIST Establishes AI Safety Consortium It is necessary for services interested in AI to stay upgraded on the policies that affect them, or they
put themselves at danger of heavy non-compliance penalties and reputational damage. Research published in March 2024 found that just 2%of big companies in the U.K. and EU knew the incoming EU AI Act. More must-read AI protection 5. Financial investment in generative AI is increasing Financing for generative AI items that generate material in reaction to a prompt almost octupled from 2022 to 2023, reaching$25.2 billion(Figure F). OpenAI, Anthropic, Hugging Face and Inflection, among others, all got
substantial fundraising rounds. Figure F Overall worldwide personal financial investment in generative AI from 2019 to 2023. Image: AI Index Report 2024/Stanford University
HAI/Quid
, 2023 The buildout of generative AI capabilities is likely to meet demand from organizations seeking to embrace it into their procedures. In 2023, generative AI was pointed out in 19.7% of all earnings calls of Fortune 500 business, and a McKinsey report revealed that 55% of organisations now utilize AI, including generative AI, in a minimum of one service system or function. Awareness of generative AI flourished after the launch of ChatGPT on November 30, 2022, and ever since, organisations have been racing to include its capabilities into their products or services. A recent survey of 300 worldwide businesses performed by MIT Innovation Evaluation Insights, in collaboration with Telstra International, found that participants expect their number of functions releasing generative AI to more than double in 2024. SEE: Generative AI Defined: How it Works, Benefits and Dangers Nevertheless, there is some evidence that the boom in generative AI”could pertain to a fairly quick end”, according to leading AI voice Gary Marcus, and services ought to be wary. This is mainly due to limitations in existing innovations, such as possible for bias, copyright issues and mistakes. According to the Stanford report, the limited amount of online data offered to train models might worsen existing issues , putting a ceiling on enhancements and scalability. It states that AI companies
might run out of premium language data by 2026, low-grade language information in 20 years and image information by the late 2030s to mid-2040s. 6. Criteria for LLM duty vary widely There is substantial variation
in the criteria that tech companies assess their LLMs against when it pertains to credibility or obligation, according to the report(Figure G). The scientists wrote that this”complicates efforts to systematically compare the threats and constraints of top AI models.”These risks include biassed outputs and dripping private info from training datasets and discussion histories. Figure G The responsible AI criteria used in the development of popular AI models. Image: AI Index Report 2024/Stanford University
HAI Reuel, a PhD student in the Stanford Intelligent Systems
Laboratory, informed TechRepublic in an email:”There are currently no reporting requirements, nor do we have robust examinations that would enable us to with confidence say that a design is safe if it passes those evaluations in the first location. “Without standardisation in this location, the risk that some untrustworthy AI models might slip through the fractures and be integrated by organizations boosts.”Designers may selectively report criteria
that favorably highlight their design’s performance,”the report added. Reuel told TechRepublic:”There are multiple reasons a damaging design can slip through the cracks. To start with, no standardised or required evaluations making it tough to compare models and their(relative )risks, and second of all, no robust evaluations, particularly of foundation models, that permit a strong, comprehensive understanding of the absolute risk of a design.”7. Staff members fidget and worried about AI The report also tracked how attitudes towards AI are changing as awareness boosts. One survey discovered that 52 %reveal nervousness towards AI services and products, which this figure had actually
increased by 13%over 18 months. It also found that just 54%of grownups agree that products and services using AI have more advantages than drawbacks, while 36%are afraid it may take their task within the next five years(Figure H). Figure H Worldwide viewpoints on the effect AI will have on present jobs in 2023. Image: AI Index Report 2024/Stanford University HAI/Ipsos, 2023 Other surveys referenced in the AI Index Report found that 53 %of Americans currently feel more worried about AI than thrilled, and
that the joint most common concern they have is
its effect on jobs. Such concerns might have a particular influence on worker mental health when AI innovations start to be incorporated into an organisation, which business leaders ought to monitor. SEE: The 10 Best AI Courses in 2024 8. US and China are developing the majority of today’s popular LLMs TechRepublic’s Ben Abbott covered this trend from the Stanford report in his short article about building AI foundation designs in the APAC region.
He wrote,
in part:” The dominance of the U.S. in AI continued throughout 2023. Stanford’s AI Index Report released in 2024 discovered 61 significant models
had been released in the U.S. in 2023; this led China’s 15 brand-new designs and France, the most significant contributor from Europe with 8 designs(Figure I ). The U.K. and European Union as an area produced 25 noteworthy models– beating China for the very first time since 2019– while Singapore, with 3 designs, was the only other manufacturer of notable large language models in APAC.”Figure I The U.S. is outmatching China and other nations in the advancement of AI designs. Image: Date Methodology The AI Index Report 2024″tracks, looks at, distills, and pictures information associated with expert system”. It makes use of a mix of data analyses, expert surveys, literature reviews and qualitative assessments conducted by worldwide researchers to supply insights into the state and trajectory of AI research study. Source
