Introduction

In the past few years, the development and expansion of artificial intelligence (AI) has resulted in many operations becoming more efficient by improving productivity in different areas. It directly impacts several firms, reducing their production time, making them more effective, and improving their production processes. Its most important and controversial outcome is that it tends to replace repetitive jobs with automation, directly impacting areas with a high concentration of Latino workers such as agriculture, service, and construction.

Generative AI and the large language models (LLMs) that power it deserves a lot of attention. They are expected to persist and become the dominant tech trend for the foreseeable future. These models generate disruption for different economic sectors and will determine how people relate with technology.

The impact that the new technology could have on Latino firms, workers, and families is huge. Latinos could be disproportionately affected by the automation of jobs, especially those currently operated manually, and it is highly concerning. A recent survey by the Pew Research Center shows that over 20% of Latino workers stated their concern about job replacement by AI. Another disturbing issue is the representation of Latino identities by AI models, often based on stereotypes that do not reflect their racial and cultural diversity, which underscores the importance of including Latino viewpoints and cultures in AI training data.

Nevertheless, despite huge concerns about AI, some Latinos with technical expertise are contributing to its development and expansion. As AI expands, it is extremely important to implement training programs on AI models and technologies for Latino firms, workers, and their families to reduce the growing gap in the use of these models between Latinos and other racial groups, especially whites and Asians.

Since AI made data more available to businesses, it has primarily been placed in large databases with limited practical use. Recent developments in generative AI are facilitating the access of decision-makers to these databases. As Amanda Kelly, co-founder of Streamlit, an open-source app framework that turns data into shareable web apps, states, “We are only starting to see how AI will assist us with cognitive work, the way robotic machinery that can lift tons helps with physical work.”

A report entitled “Data + AI Predictions 2024” by Snowflake, a data cloud company, analyzes “the impact of an AI revolution that is focused on the power of LLMs and the transformative potential of natural language interfaces. Now with computers we can talk to data like it is a person, which has huge implications for jobs, cybersecurity, data strategy, and other applications.” As a result, “generative AI is changing everything, for better and worse.”

It is expected that the AI revolution will continue for an indefinite time; currently it is impossible to stop it due to its capacity to generate new models and applications. Some corporate leaders have expressed concerns about costs and technical challenges related to its systematic expansion that could slow further development of generative AI and LLMs, but a continued expansion of both is anticipated.

According to Christian Kleinerman, executive vice president of product at Snowflake, “There is a lot of opportunity to improve things in the business world, whether it is around making individuals more productive, or creating innovative end user experiences and interactions. It will change roles, responsibilities, and skill sets.”

Industry experts have identified the following three immediate concerns that will be challenging during the early years of vast availability of generative AI and large language models.

I. An Effect on Jobs

Generative AI and LLMs could have a negative impact on some jobs, especially for people working within the so called “knowledge economy,” which requires a high degree of cognitive skill and expertise, typically in areas of problem-solving, critical thinking, and decision-making. Ethan James Whitfield, the author of many books on AI, says, “This type of work is often characterized by its intangible nature, as opposed to physical labor or manual tasks.” According to Snowflake CEO Sridhar Ramaswamy, quoted in the “Data and AI Predictions 2024” report, “Rapid change forced by widespread AI adoption would make it hard to quickly absorb displaced workers elsewhere in the workforce. ‘Both the private sector and governments will need to step up.’”

II. Deepfakes

They constitute another important problem. “In the next few years, we can expect to see an assault on what we humans collectively think of as our reality,” says Ramaswamy. “A world where no one can or should trust a video of you because maybe it was AI-generated. That is a very different reality from the one we are living in. That is a big issue.”

III. A Worsening Digital Divide

Ramaswamy is highly concerned about the long-term impact of generative AI and LLMs. “These advances will exacerbate the divide between the haves and have-nots that has been happening over the past 20-30 years,” he said, noting that advanced AI could exacerbate inequality across the globe.

At the same time, he said, “By making information so much more accessible, this technology produces a new generation of young adults who better understand the issues and potential, and can counter that risk.”

Methodology

Data used for this paper came from a recently published survey by the U.S. Census Bureau, which includes the use of AI by firms and families at the county level, although we focused on the state level. The following are the census files that contain the data: https://www.census.gov/library/stories/2023/11/businesses-use-ai.html AB1800TCB01B: Annual Business ... - Census Bureau Table.

ABS - Technology Characteristics of Businesses: 2019 Tables (Employer Businesses) (census.gov)

For this paper, we concentrated our efforts on basic statistical analysis related to the use of AI by firms in the 50 U.S. states, a cross-tabulation analysis showing the level of current AI use by race and ethnicity, and a revenue comparison by each level of AI use by Latino and non-Latino firms.

Literature Review

There are many recent publications related to the description of the characteristics and scope of generative AI and LLMs, their impact on firms’ efficiency, productivity, and revenue, the growing gap in the use of these models by different racial and ethnic groups, and their effect on minority workers, among other issues. Snowflake’s “Data and AI Predictions 2024” report includes five sections related to AI. It starts with an overview titled “The AI Era Is Upon Us,” followed by how Gen AI and LLMs will change lives and transform business, technical roles in an AI world, the cybersecurity challenges associated with AI, and key innovations that these models produce.

The AI Era is Upon Us

Noting that generative AI is “the big tech story of 2023,” the opening paragraph states, “Gen AI and the large language models (LLMs) that power it are worthy of the attention. And they are likely to remain the dominant tech trend for the foreseeable future.” It continues, “We’ve been told that AI will take our jobs, create amazing new jobs, identify diseases and discover new medicines, empower and detect deepfakes, undermine/transform education and destroy democracy, fuel and/or strangle creativity, and be our friend and our therapist . . . ”

“Generative AI and LLMs are definitely the hottest topic right now—they’re sucking all the oxygen from the room,” says Mona Attariyan, Snowflake’s director of machine learning. “It almost feels like there is nothing else in machine learning that anyone wants to talk about.”

The report discusses the potential for generative AI to revolutionize enterprise efficiency and enhance user interactions with technology, forecasting significant changes in productivity, job roles, and data strategies.

“Lots of true disruption is coming. Mostly around end user experience and how people interact with technology,” says Christian Kleinerman, noting it is not just hype. “The dramatic change everyone’s talking about is real. Generative AI and related technologies will affect productivity, job roles and responsibilities. It will aid creative processes and create entirely different experiences.” The overview concludes, “Gen AI is changing everything, for better and worse.”

Gen AI and LLMs Will Change Our Lives—Profoundly

This section predicts that AI’s impact on society will be huge and fast. “It is comparable to the arrival of the smartphone,” says Mona Attariyan. “Since the iPhone, the amount of time we spend accessing data and applications has gone through the roof, really changing how we move through our lives. The arrival of AI will be a similar step change, only much faster.” On the other hand, the report states, “Generative AI’s’s negative effects, including job loss, deepfakes, and a deepening digital divide, will be hard to manage at first.”

Gen AI and LLMs will Transform the Enterprise

Since Gen AI and LLMs are based on data management, the report considers that “For years, companies have been urged, or admonished, to develop a comprehensive and forward-looking data strategy. Just as more and more businesses were ticking that box, AI advances threaten to render last year’s plan moot. Fortunately, our experts were unanimous that if you have already put in work to create a solid data strategy, you’re on the right track.”

“The generative AI era does not call for a fundamental shift in data strategy,” says Jennifer Belissent, principal data strategist at Snowflake. “It calls for an acceleration of the trend toward breaking down silos and opening access to data sources wherever they might be in the organization.”

Technical Roles in an AI World

“LLMs and generative AI are going to have a big impact on the most technical of data users, and it is largely for the better,” the report states. “The traditional complaint in data science is that a lot of the actual work is basic data prep, boring stuff. . . . The promise of the new AI era is that basic data preparation will be automated by smarter AI tools.”

As a result, the report predicts:

1. Data engineering will evolve and be highly valued in an AI world.

2. Data scientists will have more fun.

3. Business intelligence analysts will have to uplevel.

4. Developers expect to be 30% more efficient using generative AI assistants.

Cybersecurity: The AI Challenge

“All new technologies demand that we consider their risks,” according to the report. “But AI developments are fast-moving and startling in their breadth and capability, and thus will be extra-challenging to security teams.”

The report considers that LLMs will be secured in-house. Mario Duarte, Snowflake’s VP of security, noted that his colleagues Kleinerman and James Malone, senior manager of product management, had discussed the AI supply chain that will allow businesses to construct large and not-quite-large language models within their secured environment. “That’s essential to good security,” says Duarte.

Bad data poisoning of a model is another frequent concern, particularly if an adversary deliberately introduced it. Eventually, AI data will be a target of attack. According to Duarte, “Legitimate businesses are careful about adopting and using new technologies—there’s cost, regulatory requirements and reputational risk if it is done poorly.”

Innovation as a Result of AI Models

Ramaswamy indicates some examples: i) Advancements in self-driving cars; ii) The revolution in battery design, with longer lasting batteries giving electric vehicles more range and affecting stationary power storage; iii) Biotech breakthroughs, such as the rapid development of the COVID-19 vaccine.

AI's Impact on Latinos

Other reports emphasize the impact of AI models on Latino communities. Among them we can cite:

“Understanding the Future of Latinos in the AI Era,” a report by the Aspen Institute, discusses the impact of AI on Latinos in the U.S., highlighting that the automation potential of jobs commonly held by Latinos is significant. This demographic is at risk of being disproportionately affected by job automation due to their concentration in manual and repetitive roles. It also emphasizes the importance of equipping Latinos with digital skills necessary for future employment opportunities in the AI era.

“AI Tends to Default to Latino Stereotypes,” an article published by The Drum, discusses how generative AI models, when prompted to represent Latino identities, often default to stereotypes. It highlights the broader problem of AI’s inability to grasp the full racial and cultural diversity within the Latino community, which could potentially undermine efforts toward inclusive representation.

“Considering Hispanic Americans and Artificial Intelligence,” a report by the Technology, Race, and Prejudice (TRAP) Lab, explores the specific challenges Hispanic Americans face with AI technologies. It emphasizes the importance of including Hispanic perspectives in AI training data, system classification, and development to avoid biases and enhance the user experience for this demographic group.

Another article by Axios, “AI’s Role in the Microsoft Workforce and Its Impact on Latinos and Latin America,” is based on a survey performed by Microsoft’s Annual Work Trend Index, which reveals that Latin Americans and U.S. Latinos are more open than other groups to using AI for tasks like summarizing meetings, despite concerns that AI might replace their jobs. Over half of the U.S. Latino workers expressed worries about job replacement by AI, a sentiment shared by half of the respondents in Latin America. This information comes as tech companies continue to enhance AI accessibility, including language expansions in AI tools like Google’s Bard and OpenAI’s ChatGPT.

The Spanish newspaper El País published an article that highlights the significant contributions of Latino technologists to the AI revolution, focusing on efforts to ensure technology serves diverse communities. Leaders like Laura Montoya and organizations such as LatinX and AI4ALL are working to make AI education accessible, promote diversity in AI development, and address social issues through technological development.

Data Analysis

We performed a brief statistical analysis using data published by the U.S. Census Bureau (USCB) on the use of artificial intelligence at different levels: high, low, and moderate, by race and ethnicity of the firm owners. Table 1 shows the total number of Hispanic and non-Hispanic firms that participated in the survey at the three considered levels of AI usage. Table 1 also includes the number of participating firms that stated they were not using artificial intelligence in their operations and the total number of Hispanic and non-Hispanic firms that participated in the survey.

Table 1. Number of Hispanic and Non-Hispanic Firms Using Artificial Intelligence

Source: https://www.census.gov/library/stories/2023/11/businesses-use-ai.html AB1800TCB01B: Annual Business ... - Census Bureau Table

Technology	Hispanic Firms	Non-Hispanic Firms	Total Number of Firms
AI: High Use	3,813 (0.78%)	83,728 (1.01%)	87,541 (0.99%)
AI: Moderate Use	6,585 (1.35%)	139,367 (1.68%)	145,952 (1.66%)
AI: Low Use	7,252 (1.48%)	35,284 (0.42%)	42,536 (0.48%)
AI: Did not use	471,766 (96.39%)	8,055,830 (96.9%)	8,527,596 (96.87%)
AI: Grand Total	489,436 (100.0%)	8,314,209 (100.0%)	8,803,645 (100.0%)

 

Table 1 shows the still very low usage of artificial intelligence by Hispanic and non-Hispanic firms in the three considered categories. The percentage of firms reporting a high use of artificial intelligence is on average close to 1%, with only 0.78% of Hispanic firms reporting high usage. More Hispanic and non-Hispanic firms report moderate levels of AI usage. The number of moderate users of AI reported by Hispanic firms was 2,772 higher than those reporting high use; the number of non-Hispanic firms reporting moderate use was 55,639 higher than those reporting high use. The percentage of moderate AI usage for both types of firms also was higher, at 1.68% for non-Hispanic and 1.35% for Hispanic-owned firms. 

It is interesting to note that at the low level of AI usage, the number and the corresponding percentage of Hispanic-owned companies increased with respect to those reporting moderate usage of AI. A low AI usage level by Hispanic firms reached close to 1.5% of all participating Hispanic firms. On the other hand, non-Hispanic firms presented an inverse trend with their number and percentage declining from moderate to low usage of AI. The percentage of non-Hispanic firms went from 1.68% for moderate usage to 0.42% for low usage. This data shows that AI usage is still low among all types of firms, although the numbers and corresponding percentages are steadily increasing.

Table 2 shows that only 5.56% of all firms participating in the USCB survey were Hispanic-owned companies. At the time of the survey a little over 25% of Hispanic companies were already reporting some use of artificial intelligence. The relatively low level of usage of AI by Hispanic companies could be explained in part by the main constraints these firms face in adopting AI—its high cost and, in many cases, the firms’ lack of appropriate technological expertise. Over 17% of participating Hispanic firms reported only a low use of artificial intelligence in their operations.

Table 2. Percentage of Hispanic and Non-Hispanic Firms Reporting Usage of Artificial Intelligence

Source: https://www.census.gov/library/stories/2023/11/businesses-use-ai.html AB1800TCB01B: Annual Business ... - Census Bureau Table

Technology	Hispanic Firms	Non-Hispanic Firms	Total Number of Firms
AI: High Use	4.36%	95.64%	100.0%
AI: Moderate Use	4.51%	95.49%	100.0%
AI: Low Use	17.05%	82.95%	100.0%
AI: Did not use	5.53%	94.47%	100.0%
AI: Grand Total	5.56%	94.44%	100.0%

Table 3 shows the percentage of firms of the different racial groups that are using artificial intelligence in their operations. As expected firms predominantly owned by whites have the largest percentage of use at the three different levels, followed by Asians, Hispanics, and Blacks. Among all firms that reported the use of AI at any of the three considered levels, we found that close to 84% of white-owned firms are somehow using AI, while only 1.8 percent of Black firms are using it. Close to 9% of Asianowned firms and 5.3% of Hispanic-owned firms were using AI at the time of the survey.

The largest percentage of Hispanic-owned firms using artificial intelligence is at the high use level, with close to 9%; for whites, it is at the moderate use level with a little over 80%. The higher percentage usage of AI for Asian-owned firms is at the low use level, with close to 12.5%. The usage of AI by Black-owned firms is at the lowest across the board among the racial and ethnic groups included in the survey.

Table 3. Percentual Usage of Artificial Intelligence of Firms by Race

Source: https://www.census.gov/library/stories/2023/11/businesses-use-ai.html AB1800TCB01B: Annual Business ... - Census Bureau Table

Technology	Hispanic	Asian	Black	White	Total
AI: High Use	8.95%	11.19%	1.93%	77.93%	100.0%
AI: Moderate Use	6.81%	10.79%	2.77%	80.04%	100.0%
AI: Low Use	4.72%	12.47%	2.77%	79.64%	100.0%
AI: Did not use	5.22%	8.87%	1.80%	84.11%	100.0%
AI: Grand Total	5.25%	8.95%	1.83%	83.97%	100.0%

The percentage difference in the moderate use of AI between white- and Asian-owned firms, the groups with the most users at the moderate level, is 69.25; the difference in low use of AI between white- and Black-owned firms, who reported the lowest usage, reaches 76.87%; and finally between white and Hispanic firms, the difference is 74.92%. Table 3 shows the huge gap in the usage of AI by the different racial and ethnic groups, which needs to be addressed to make the development and expansion of artificial intelligence more sustainable.

Figures 1 and 2 are a graphical representation of the data included in Table 3. Figure 1 shows the enormous gap between the use of AI by white-owned firms versus minority-owned firms. It is very interesting to note that among all firms that are not using AI, white-owned firms also reported the largest numbers, reaching over 84%. This could be explained in part by the large number of white-owned firms that responded to the survey.

If we consider only white-owned firms that participated in the USCB survey in large numbers compared to minority-owned firms, we can observe that most reported using artificial intelligence at moderate (80%) and low (79.6%) levels, followed by 77.9% reporting high use of artificial intelligence. This clearly shows that even white-owned firms still are using AI primarily at moderate and low levels.

Source: https://www.census.gov/library/stories/2023/11/businesses-use-ai.html AB1800TCB01B: Annual Business ... - Census Bureau Table

Figure 2 shows a comparison in the usage of artificial intelligence among minority-owned firms that participated in the USCB survey. We decided to show a graph on the usage of artificial intelligence only by minorityowned firms due to the huge gap in the use of artificial intelligence between these firms and white-owned firms. This graph illustrates differences in the usage of artificial intelligence by the three selected racial/ethnic groups in the three considered levels of usage: high, moderate, and low. In addition, we include the “did not” category

Source: https://www.census.gov/library/stories/2023/11/businesses-use-ai.html AB1800TCB01B: Annual Business ... - Census Bureau Table

Asian-owned firms reported the highest usage of artificial intelligence among the three considered minority-owned firms. About 12.5 percent of these firms reported low use of AI, followed by high use at 11.2%, and moderate use at 10.8%. Only 8.9% of Asian-owned firms that participated in the survey were not using AI.

It is interesting to note that among Hispanic-owned firms using AI, nearly 9 percent reported high use, followed by moderate use at 6.8 percent, and low use at 4.7 percent. These figures clearly show how interested Hispanic-owned firms are in using AI, although their current usage is lower than that of white-owned firms, resulting in a 79 percent gap.

Black-owned businesses report the lowest usage of AI in all considered levels in Figure 2. These firms need a lot of information, training, and financial support to increase their AI usage levels.

Table 4. Percentage Increase of Revenue Generated by the Usage of Artificial Intelligence by Race and Ethnicity.

Source: https://www.census.gov/library/stories/2023/11/businesses-use-ai.html AB1800TCB01B: Annual Business ... - Census Bureau Table

Technology	Hispanic	Asian	Black	White
AI: High Use	3.54%	8.65%	0.64%	91.83%
AI: Moderate Use	2.97%	6.54%	0.49%	90.99%
AI: Low Use	2.30%	5.23%	0.45%	87.35%
AI: Did not use	2.90%	5.18%	0.77%	91.15%
AI: Grand Total	5.25%	8.95%	1.83%	83.97%

According to data included in Table 4, participating firms in the USCB survey reported a direct impact on their revenue by the usage of artificial intelligence. As expected, the percentage of revenue generated by low use of AI is the lowest. It increases for firms with moderate use, and reaches the highest level when firms have a high use of AI.

The gap for Hispanic-owned firms between those with high use and those with low use of AI is 1.24%, for Asian-owned firms the gap is 3.42%, and for Black-owned firms it is 0.19%.

The percentage of revenue generated by use of AI for white-owned firms mirrors the trend described above for minority-owned firms. Firms reporting low use of AI report revenue generated equal to 87.4 percent; it increases to 91.0 percent for those reporting a moderate use of AI and reaches 91.8 percent for those reporting a high use of AI. These figures clearly show the importance of AI usage by all types of the considered firms since it has a direct impact on their revenues, by making their processes more productive and efficient.

Table 5. Percentage Increase of Revenue Generated by the Usage of Artificial Intelligence by Race and Ethnicity.

Source: https://www.census.gov/library/stories/2023/11/businesses-use-ai.html AB1800TCB01B: Annual Business ... - Census Bureau Table

Technology	Hispanic Firms	Non-Hispanic Firms
AI: High Use	$4,472,110 (3.70%)	$120,854,280
AI: Moderate Use	$13,796,728 (3.14%)	$439,976,606
AI: Low Use	$15,226,700 (2.41%)	$632,691,532
AI: Did not use	$655,952,296 (3.10%)	$21,154,013,594
AI: Grand Total	$689,447,834 (3.09%)	$22,347,536,012

Table 5 shows the average monetary value of the revenue reported by participating Hispanic and non-Hispanic firms reporting no, low, moderate, and high levels of artificial intelligence usage. We can observe that the revenue generated by all Hispanic firms (grand total) that participated in the USCB’s survey represents only 3.09 percent of the revenue generated by all other firms. This difference could be explained by the small size of some Hispanic firms, the industrial sector under which they are categorized, the price of their products, the automation of their operations, and the geographical location of their main offices and operations.

Conclusion

Artificial intelligence is here to stay and its uses and applications will continue to expand. It could have positive and negative impacts on families and businesses in general, making production processes more efficient, shortening the time between when an idea emerges about a certain product and its inception in the market. AI facilitates the analysis of huge data sets, with the potential to replace tedious jobs with automation and streamline the decision-making process for top administrators.

In industrial sectors such as agriculture, service, and construction, AI has the potential to use robotics to automate repetitive tasks. On the one hand, this shift can make firms more efficient by reducing their production time and costs. On the other hand, there is a high concentration of Hispanic workers in these sectors who are concerned about losing their jobs.

AI could have many negative impacts to individuals and firms due to deepfakes, which can be used to spread misinformation and share AI-generated false videos of people on social media.

Another huge potential negative is the introduction of false data. Since AI is generated from large data sets, it would be very difficult to differentiate which data set is real or bad. The use of these false data sets could result in outcomes that could cause serious problems for users.

Finally, it is essential to emphasize how important it is for the Hispanic community to train and educate themselves on how to use AI for their benefit. Since most of them are currently working in areas with highly repetitive jobs that could be eliminated by AI, it is necessary to start preparing them to work in other industrial sectors.

---