Connect with us

Business

Businesses love AI — but so do hackers

A look at the top cyber threats to machine-learning systems, according to Google’s AI red team.

Published

on

Share this:

You can’t generate impressive results, efficiency, and new solutions without a little risk. And in the world of digital transformation, cybersecurity is the main risk to look out for. We’ve seen this in agricultural tech advancements, the IoT, and even autonomous vehicles

AI, specifically machine learning, is no exception. 

As developers and engineers continue to safeguard machine-learning systems with updated security measures, cybercriminals continue to evolve their tactics to penetrate those systems. 

Google recently launched a dedicated team of cybersecurity professionals to study this very topic — i.e. a red team. Specifically red teams are trained to think like a hacker so they can accurately predict and combat hacker tactics. Or, as Wikipedia describes them, “a group that pretends to be an enemy.”

In an interview with The Register, the head of Google Red Teams Daniel Fabian discusses common cyberattacks businesses with machine-learning systems need to look out for:

  1. Data poisoning

Machine-learning models need to learn before they can function. And it’s in that training stage that they’re most vulnerable. Today’s cybercriminals can alter (poison) the date in those training models to change the function of a machine-learning system. 

“Anyone can publish stuff on the internet, including attackers, and they can put their poison data out there. So we as defenders need to find ways to identify which data has potentially been poisoned in some way.”

  1. Prompt injection attacks

Hackers can also tinker with a language learning model’s (LLM) output. This usually entails code to instruct the model to ignore previous instructions, and the code will provide new commands that can switch the intended action to a more nefarious one.  

  1. Backdoor 

Like the name suggests, a backdoor cyberattack entails creating a hidden entry to the model’s code. Keyword hidden — hackers can move into the model’s code and bypass any implemented authentication measures. 

“On the one hand, the attacks are very ML-specific, and require a lot of machine learning subject matter expertise to be able to modify the model’s weights to put a backdoor into a model or to do specific fine tuning of a model to integrate a backdoor.”

  1. Adversarial attacks

Hackers can feed specialized inputs into a machine-learning model and lead it to make mistakes or produce incorrect outputs. 

But while we must remain vigilant, there’s no need to panic. Fabian predicts that it will get easier for cyber professionals to predict weaknesses and vulnerabilities and thus protect their data and machine-learning systems:

“In the long term, this absolutely favors defenders because we can integrate these models into our software development life cycles and make sure that the software that we release doesn’t have vulnerabilities in the first place.”Read the full article on the Register here.

Share this:

Business

Americans' pandemic-era entrepreneurial streak is holding strong—for now

Published

on

By

An altLINE analysis of Census Bureau data reveals Americans are still starting new businesses at higher rates than in pre-pandemic times.
Share this:

Inflation has nothing on the American entrepreneurial spirit, which, judging by the volume of new businesses formed, continues to see potential in the post-pandemic economy.

To better understand the post-COVID-19 outlook for entrepreneurship in the U.S., altLINE analyzed data and reports from the National Bureau of Economic Research and the Census Bureau. The data shows that Americans are on track through July of this year to submit 54% more applications to start new businesses compared to the same period in 2019, before the onset of the pandemic.

New business applications soared initially at the start of the COVID-19 pandemic as brick-and-mortar businesses were forced to close their doors in compliance with local social distancing mandates. Stores saw business plummet and many were never able to reopen their doors, even as public health restrictions eased. The seismic shift in shopping habits spurred many Americans to start new business ventures at rates not seen since before the Great Recession, when the U.S. consumer took a hit from one of the deepest recessions on record.

As the current economic situation puzzles economists who debate whether a recession may be in the future, the continuing creation of businesses could mitigate some of the pain of a slowing economy.

Studies have suggested that the growth of the smallest businesses can help an economy’s resilience. Young, tiny companies, sometimes called “microbusinesses,” reduce local unemployment rates in their communities and have been related to rising household incomes, according to GoDaddy’s July 2021 Venture Forward Report.


A line chart depicting business formation applications submitted every month from 2019-2023. The trend line spikes in 2020, comes down slightly and then continues growing slowly in 2023.

altLINE

COVID-19 recession provides shot in the arm

Advancements in technology made it easier for business owners to set up and run online storefronts and services. Leading up to 2020, ecommerce platforms integrating new technology for enhanced shopper experiences provided a critical foundation for the spike in new businesses. As Americans stayed at home during the height of the pandemic, they shopped online for everything from personal care to groceries. Ecommerce sales nearly doubled in 2020, jumping by 43% to a whopping $815 billion in annual retail sales. Thousands in stimulus checks also did their part to keep Americans afloat—and spending. On top of those factors, interest rates for loans to buoy new companies and purchase real estate were at historic lows.

In the first year of the new business surge, retailers in the fashion space made up the lion’s share of new small businesses, according to the GoDaddy Venture Forward Report.

Today, those new business owners face a much more expensive economy. Costs for labor, gas, clothing, food, and other critical inputs for businesses have risen considerably since 2020.

Business owner using mobile app on smartphone checking a parcel box.

Ground Picture // Shutterstock

New business class faces considerable headwinds

The typical new business faces its most difficult time in its first years of operation. Historically, 4 in 5 new businesses make it beyond their first year, according to Bureau of Labor Statistics data. But the odds of survival dwindle in each subsequent year of operation. Based on trends, just 1 in 2 businesses created in 2020 will likely survive beyond 2025.

The entrepreneurs looking to survive now face mounting headwinds in the face of rising interest rates, which has made borrowing money more expensive for both consumers and small business owners.

For small businesses seeking venture funding, seed-stage venture capital has stagnated as the venture capitalist ranks have grown wary of investing in early-stage companies. For those seeking loans, the cost of borrowing money today is at its highest since 2001, when the tech bubble burst, throwing the U.S. into recession.

Story editing by Ashleigh Graf. Copy editing by Kristen Wegrzyn.

This story originally appeared on altLINE and was produced and
distributed in partnership with Stacker Studio.

Share this:
Continue Reading

Business

Robots are starting to deliver takeout orders. Are they here to stay?

Published

on

By

Task Group analyzed the state of autonomous delivery systems, both nationally and internationally, to see how far along this technology has come.  
Share this:

In a March 2023 Deloitte survey, 47% of Americans said they would order from a restaurant that delivers food with a drone or an autonomous vehicle. That’s up 3 percentage points from the company’s 2021 survey about restaurant trends.

In that first survey, researchers noted there was “massive uncertainty in the industry, and many worried that restaurant patronage might never recover” from the COVID-19 pandemic. It found that two-thirds of consumers believed they would not immediately return to their pre-pandemic restaurant habits.

In 2023, most restaurant customer behavior is back to normal—though some changes have blended into the industry’s practices. Task Group analyzed the state of autonomous delivery systems, both nationally and internationally, to measure the progress of this technology post-pandemic.

As with other industries, technology has helped maximize efficiency and improve customer satisfaction. Business owners learned new service methods, marketing strategies, and technical terminology. Food delivery skyrocketed during lockdowns, making greater strides in restaurant efficiency and, in some cases, profits. Many restaurant owners connected apps that allowed customers to order without talking to a human to state-of-the-art delivery systems that don’t require a driver.

Restaurants and transportation companies in North America and Europe are experimenting with new automated delivery techniques that can reduce their costs as long as they do not compromise customer satisfaction. And consumers are ready—but how soon will it become standard practice?


Food delivery robots on pathway.

Julija Sh // Shutterstock

What are drones and sidewalk delivery vehicles?

The robots most commonly used in the food delivery industry are aerial drones and wheeled autonomous delivery vehicles that travel along sidewalks to reach customers.

Drones are classified by how they generate lift—with fixed wings, rotors, or a combination—by how they’re used, such as food delivery, and what equipment they have on board, including batteries and cameras.

In the U.S., the Federal Aviation Administration regulates drone use. The agency requires pilots to be certified—and bans drones within five miles of airports.

For many years, the FAA stood in the way of companies seeking to use drones for deliveries, but in 2019, the agency agreed to allow uncrewed delivery flights beyond the pilot’s line-of-sight by UPS and Wing Aviation, owned and operated by Google’s mothership Alphabet. Since then, the agency has approved drone delivery operations for several companies, including Amazon and Walmart.

According to a study published by the Harvard Kennedy School in 2022, autonomous delivery vehicles are not the future. They’re already here. Self-driving machines about the size of a large cooler are already traveling down our sidewalks and crosswalks to deliver various packages.

Policymakers question how these vehicles will work and interact with people and other vehicles in already congested and chaotic urban environments. The Harvard researchers believe these vehicles “offer the promise of less congestion and greener shipments,” but also “raise concerns about safety and use of road and sidewalk infrastructure.”

While the debate continues, the manufacturers of these robots continue to advance their technology, including using machine learning to improve navigation, efficiency, and safety.

Food delivery drone in flight.

Canva

How far off are drone or sidewalk deliveries?

Estonia-based delivery startup Bolt, working with Starship Technologies, has been trialing sidewalk deliveries in Estonia, the U.K., and the U.S. and plans to formally launch robot deliveries later this year in as many as 500 cities in 45 countries.

Bolt’s main competitor, Uber, signed a deal in 2022 with autonomous vehicle startup Nuro “to test driverless food deliveries” in Mountain View, California, and Houston, Texas. Before the agreement, Uber ran a pilot program for sidewalk delivery in Los Angeles, while Nuro delivered Domino’s pizzas in specific areas of Houston for a year. 

Story editing by Jeff Inglis. Copy editing by Kristen Wegrzyn.

This story originally appeared on Task Group and was produced and
distributed in partnership with Stacker Studio.

Share this:
Continue Reading

Business

AI “superusers” seek education, fun, and productivity with generative AI

A look at two separate studies by Sparktoro and Salesforce on people’s generative AI use.

Published

on

Share this:

Maybe it was through your job. Or simply out of curiosity.

With the rise of generative AI, you’ve probably tried out ChatGPT or a similar tool. But how often are people using these? More interestingly, what motivates them? Both Salesforce and SparkToro sought to find out with two separate studies. 

Here are highlights from each report and how they compare:

Work automation and educational pursuits top priorities for AI users

Both Salesforce and SparkToro can agree on this. SparkToro highlighted professional use of the platform as at an “all-time high,” then ranked categories of interest across over 4,000 ChatGPT prompts with these in the top 5:

  • Programming: 29.14%
  • Education: 23.30%
  • Content: 20.79%
  • Sales and Marketing: 13.47%
  • Personal & Other: 6.73%

Salesforce found that 75% of generative AI users are motivated by streamlined work communications and task automation. The second highest topic of interest? Technically “messing around” (38%), though a close third was learning and education (34%). Both SparkToro and Salesforce posit that education doesn’t just include homework or university coursework—users also use tools like ChatGPT to develop knowledge of other desired educational topics. 

Younger generations more likely to use AI than older ones despite general decline in usage

Salesforce surveyed 4,000 people to find out how they use generative AI and what their demographics are. Turns out, most “superusers” — aka those who use the tool every day — are Millennials or Gen Zers (65%). Plus, 70% of the Gen Z participants surveyed said they use generative AI. 

Still, SparkToro notes an overall decline in generative AI use regardless of age. After studying monthly traffic data on OpenAI provided by Datos, SparkToro found overall traffic fell by nearly 30%. 

Users ask ChatGPT to write, create, and list

These were the top three common words in SparkToro’s assessment in ChatGPT prompts. However, they also share a notable prevalence of the words “game” and “SEO in prompts as well. Other words less commonly used yet enough to come up in the results included judge, SaaS pricing, curriculum, employment, and employer.

Read the SparkToro report here and the Salesforce report here

Share this:
Continue Reading

Featured