Desktop version

Home arrow Psychology

  • Increase font
  • Decrease font


<<   CONTENTS   >>

Synthetic Colleagues

We humans have long envisioned artificially intelligent helpmates. Science fiction is replete with visions of these (some more helpful than others). Notables include: Robby from the movie Forbidden Planet, Rosie from The Jetsons, Colossus from the movie Colossus: The Forbin Project, the T-800 (Model 101) from the Terminator series, Data from Star Trek: The Next Generation, KITT from the television show Knight Rider, Andrew from Bicentennial Man, JARVIS from the Ironman series, Samantha from the movie Her, and HAL from the classic 2001: A Space Odyssey.

Neither cognitive systems nor artificial intelligence are advanced enough to approach these fictional systems yet but progress is being made with recent significant achievements described elsewhere in this book. Until such artificial systems are possible, humans will work with and collaborate with cognitive systems. Cogs will be our partners working alongside us and achieve Level 3 and Level 4 cognitive augmentation as described in Chapter 3. The researcher/cog collaboration at the professional level forms a synthetic colleague with the output being the emergent result of biological and artificial thinking.

Our collaborative view of cognitive systems mirror's IBM's view (Kelly and Hamm, 2013):

.. .humans and machines will collaborate to produce better results—each

bringing their own superior skills to the partnership...

In 1987, Apple, Inc. envisioned a collaborative intelligent assistant called the Knowledge Navigator (Apple, 1987). The Knowledge Navigator was a vision of an artificial executive assistant capable of natural language understanding, independent knowledge gathering and processing, and high-level reasoning and task execution. In 2014, IBM released a video demonstrating humans collaborating with an advanced version of the IBM Watson technology having won the challenge against two human Jeopardy! champions in 2011 (Gil, 2014). Some aspects of the video are strikingly similar to the Knowledge Navigator video of 1987, particularly the collaborative nature of the dialog. In the video, two humans collaborate with Watson on a business analysis task.

The cogs we envision will interact with us at a personal level throughout the day through a variety of interactivity mechanisms, including natural language, helping us in every aspect of our lives including our professional endeavors. Cognitive colleagues will be able to do some of the high-level thinking for us just like a human colleague. As other humans would do, cognitive colleagues will build a history and understanding with us over time, and come to know us as well as, or better than, our human co-workers, spouses, and family members. Our intellectual achievements will become a collaborative effort between our cogs and us. This makes cogs very valuable going forward into the future. They will carry an intimate knowledge and understanding of us and our achievements long after we are dead.

Enhancing Productivity

We have already entered into the era of the AI as a co-worker and the era of Al-enhanced productivity (Katz, 2017). A number of tools, varying widely in capability and sophistication, exist playing three typical roles in the work environment (Chu and Wang, 2019):

  • • Automating business processes
  • • Augmenting business decision-making
  • • Facilitating engagement with customers and other employees

Automation usually involves tasks such as recognizing entities, detecting patterns, classification, extracting information, and search. Automating these kinds of tasks enables companies and organizations to deliver relevant work faster, more accurately, and at a lower cost. Decision- Making uses pattern, association detection to derive actionable insights from enormous data stores at superhuman scale, speed and accuracy. The cognitive insights are useful to executives because they improve the quality of strategic decisions. In customer and employee engagement, digital agents enhance the customer experience by answering questions contextually based on past behavior, preferences, weather, vacation plans, etc. In finance, digital advisers give investment advice, provide personalized investment options to customers, and monitor and alert users about changes to portfolio risk. In healthcare, digital agents deliver personalized medical advice to patients.

For its coverage of the 2016 Olympics in Rio, the Washington Post developed an artificial story editor named Heliograf. According to Jeremy

Gilbert, director of strategic initiatives at The Washington Post "In 2014, the sports staff spent countless hours manually publishing event results. Heliograf will free up Post reporters and editors to add analysis, color from the scene and real insight to stories in ways only they can" (WashPostPR, 2016). Heliograf automatically started hundreds of stories which were finished by human editors resulting in an increase in output of several hundred percent.

In 2019, OpenAI announced a language model called GPT-2 able to predict the next word in a block of text. The result of unsupervised machine learning, and trained on a dataset of 8 million Internet pages, GPT-2 has a broad set of capabilities, including the ability to generate conditional synthetic text samples of unprecedented quality. In fact, GPT-2 is so good OpenAI is refusing to release the full model to the public for fear of malicious use. GPT-2 also is able to answer questions, display reading comprehension, summarize, and translate (Radford, 2019). Although GPT-2 falls short of these higher-level skills, it represents an important new way these capabilities can be self-learned possibly leading to cogs able to learn and improve rapidly to superhuman levels.

At Fennemore Craig, an Arizona-based corporate law firm, lawyers use a system developed by ROSS Intelligence to comb through millions of pages of case law and write up findings in a draft memo. The process, which might take a human lawyer four days, takes ROSS roughly 24 hours. ROSS doesn't suffer from exhaustion or burnout: The tool can pull infinite all-nighters without its work suffering as a consequence (ROSS, 2019).

Scriptbook is a cloud-based system able to read a movie screenplay and predict the film's MPAA rating, the gender and race of the target audience, and box office performance in only a few minutes. ScriptBook has been trained on a dataset of 6,500 existing scripts. Currently, film studios rely on subjective human evaluation and take on a lot of risk when deciding to put a movie into production. Mitigating this risk by being better able to identify a blockbuster is worth billions of dollars to the film industry. In 2016, Scriptbook analyzed 50 scripts evaluated by humans and made into movies of which only 18 were financially successful. Therefore, human analysis and decision-making achieved a success of only 36%. However, Scriptbook correctly predicted the performance of 40 of the movies, an 80% accuracy (ScriptBook, 2019).

A similar company, StoryFit, uses cognitive system tools to make studio marketing easier and more targeted. StoryFit Comps automatically generates comp lists based on metrics including budget, genre, cast, themes, story arc, tone, and more. StoryFit MetaData makes it easy to search for films including certain topics, themes, genres, scenes, and more. Metadata makes it simple to rediscover and re-market old titles for renewed sales. StoryFit Insights yields dozens of metrics detailing the film's screenplay, including emotional content, story arc, character personalities, estimated budget, and more (StoryFit, 2019).

Fifteen years ago, translators could expect to earn about $175 a day for translating some 2,000 words. Today, working in tandem with cogs, a translators achieve 10,000 words per day, a 5-fold increase in productivity. The process, known as post-editing machine translation (PEMT), involves letting the cog take the first pass, and then bringing in a human translator in to tidy up the language, check for improperly interpreted terminology, and make sure the tone, context, and cultural cues of the translation are all on point. As Miranda Katz, associate editor at Backchannel, puts it "These AI tools are like plucky young assistants on steroids: They're highly competent and prolific, but still need a seasoned manager to do the heavy intellectual lifting" (Katz, 2017).

The human resources (FIR) field, is employing chatbots, intelligent assistants, and predictive analytics. One third of HR managers' time involves seeing candidate search results not matching the context of what they are looking for. Intelligent search capabilities can interpret a recruiter's simple keyword searches and apply custom ontology and taxonomy structure to better understand intent. The cog not only understands what is said, but what is meant. The result is more accurate search results as well as better handling of acronyms, synonyms and related concepts (Bell, 2018).

Chemists are using deep neural network machine learning to predict macroscopic molecular dynamics from the quantum mechanical wavefunction of the atoms in the molecule (Schutt et al., 2019). This opens promising avenues to perform inverse design of molecular structures for targeting electronic property optimization and a clear path towards increased synergy of machine learning and quantum chemistry.

Artificial Entertainers, Lawyers, Politicians

Langley (2013) challenged the cognitive systems community to develop an artificial entertainer, an artificial attorney, and an artificial politician to drive future research on integrated cognitive systems. Although an artificial lawyer does not yet exist, a number of legal bots do exist allowing a lawyer, or even a layperson, to interact with the trained software and find answers to legal questions, and be guided through the completion of legal forms and documents, or perform other legal processes.

A2JAuthor (https://www.a2jauthor.org/) is a cloud based software tool enabling self-represented litigants to rapidly construct legal documents. Berkeley Bridge (https://www.berkeleybridge.com/ value/) is a document-building tool driven by an expert system. Bryter's knowledge and decision automation tool (https://bryter.io/) has been used by several legal firms to enhance client services. Josef's legal automation products (https://joseflegal.com/) allow lawyers to instantly generate personalized letters and contracts. LawDroid (https://lawdroid. com/) is a chatbot automation company with technology allowing legal firms to create chatbots to automate certain activities within a legal office for convenience and to improve productivity. Neotalogic (https:// www.neotalogic.com/) offers legal service automation tools, such as PerfectNDA, so legal firms can provide a superior experience at a lower cost for their clients.

Recently, a company called BlackBoiler was issued four patents involving its artificially intelligent contract review products (BlackBoiler, 2019). The BlackBoiler system suggests revisions based on edits to previously reviewed contracts, deploys the company's legal playbook during each new contract review, and grows smarter and increases efficiency with each additional use. BlackBoiler can review and comment on a contract in less than a minute, a process requiring many hours of human effort. While not the complete artificial lawyer Langley envisioned in 2013, BlackBoiler represents the cog era in which cogs will perform more and more of the cognitive processes of a human expert in the field.

Whereas the ultimate goal of artificial intelligence is to create an artificial entity with the same range of abilities as a human, we think the near-term goal should be to create a cognitive system capable of some expert-level performance in each field of endeavor—but not necessarily total performance. A person, even someone other than a lawyer, collaborating with any of these tools forms a human/cog ensemble and therefore represents a synthetic lawyer. As the capabilities of these cogs increase, more and more of the lawyer's profession will be automated. Eventually, technology will produce a truly artificial lawyer but until then lawyers and average people will achieve Level 3 and Level 4 cognitive augmentation by working with legal cogs.

 
<<   CONTENTS   >>

Related topics