5 key areas for artificial intelligence to overcome for “difficult tasks”.

We are speeding down the highway, well above the speed limit, moving freely on the smooth road, trusting our own TeslaAutopilot can handle slight turns.

However, when we enter the Golden Gate Bridge towards San Francisco, the lane narrows, construction works and temporary barriers encroach on the left, the road becomes unpredictable, uneven and stuffed with cracks – I immediately take control of the steering wheel.

(*5*)

Anyone who has tried Autopilot in a Tesla has likely had a similar experience. It copes well enough with monotonous long stretches of highway. But when the driving gets demanding, it clearly is not up to the task.

But what’s flawed with this photo? Is there some inherent reason why AI must be good at easy, mindless tasks but bad at difficult ones? Logically, as computing power increases and models improve, at a certain threshold the autopilot should outperform the human.

In a few years, when the road narrows and driving becomes harder, I expect I’ll switch to autopilot.

AI easy task

If we glance at the state of AI firms today, we discover similarities. The world is shocked by surprising abilities ChatGPTlatest firms appear every day purporting to handle cutting-edge cases across all industries.

In most cases, these use cases are analogous to driving on a smooth, paved highway. These primarily belong to the “easy task AI” scenarios: high volume, low criticality and, compared to other scenarios, high error tolerance. Most AI-based tools today focus on repetitive tasks, enabling humans to handle mission-critical, high-value situations.

Some examples: More than $1 billion in enterprise capital has been raised for customer support AI startups, including Unifor, Forethought, Move the guns, Follow.AI AND Gorgiasin addition to products from leading manufacturers akin to IntercomAI bot, Finn.

Enterprises can tolerate errors in support workflows so long as high volumes of requests might be handled cost-effectively; customers will at all times transform into humans if the AI ​​fails to complete the task.

The status of AI in law is similar: Don’t pay is a great example of using AI to brute-force legal problems that may sometimes get you flawed, akin to fighting parking tickets, canceling subscriptions, and a host of other relatively low-stakes legal tasks.

Similarly in medicine, firms akin to Restrict save time by automating clinical notes. In this case, artificial intelligence does not replace doctors, but only increases their effectiveness.

But is this the future? It seems strange to relegate artificial intelligence models – trained on more data than an individual can ever internalize – to only relieving humans from tasks that the majority would consider repetitive, mundane, or “easy.”

Instead, it seems much more likely that AI will start taking on difficult tasks.

Artificial intelligence for difficult tasks

What does a world seem like where AI is at its best at the most difficult, mission-critical, and high-risk tasks?

We can already see the clues. Scientists from WITH AND Massachusetts General Hospital develop an AI model analyzing computed tomography, potentially detects lung cancer years sooner than a radiologist. Will there come a day in the future when a bot with a vast amount of information about previous cases shall be your primary care doctor?

The US Air Force demonstrated X-62A Vista, an AI-controlled fighter with faster response times and greater precision than a human pilot.

Defy.vc portfolio company Aircover.ai builds a virtual sales engineer: Imagine that a sales representative knew on day one how to flawlessly answer all questions that will require the presence of a product or SE expert? Would you ever conduct a high-stakes sales call without your AI assistant?

As artificial intelligence takes on tasks beyond human capabilities, here are some industries we will expect to impact:

Medicine: Artificial intelligence systems that analyze clinical data and predict diagnoses with greater precision than human doctors, and provide recommendations for medications and treatment plans. Artificial intelligence-assisted surgery is in its infancy. Drug discovery is already being revolutionized by artificial intelligence.

Transport: Autonomous vehicles are just the starting. Beyond combining trucks in a convoy, imagine that vehicles on the road communicate with each other and form a network, operating effectively as a single organism, adaptively minimizing congestion and safely moving at higher speeds, slightly than increasing delays as each driver responds.

Corporate SaaS service: Why does every CIO have to reinvent the wheel in their organization? Internal systems shall be self-integrating; automation will connect systems from different vendors to perform complex tasks. Intelligent data evaluation will happen via conversational interfaces.

Security: Advanced models already detect fraud by analyzing patterns in thousands and thousands of transactions, a feat that goes far beyond any human skill. Advanced artificial intelligence systems already assess risk among a whole bunch of 1000’s of employees. We will live in a passwordless future where enterprise security systems operate silently behind the scenes, adapting to signals from a common security network across enterprises.

Employees of the future: Questions are inevitable. What is the role of man in this latest world? What are the interfaces between humans and latest intelligent systems? What management models are needed? How much autonomy do we give to these systems? What skills should we teach our youngsters to prepare for a future in which intelligent systems can cope with difficult mental challenges?

I’d love to hear what AI use cases you envision for the future.