CRM giant Salesforce today said that it would commit $250 million to investments in startups focused on generative AI, even as the company warned of the dangers of the technology.

The company emphasized the potential gains for application software possible through the use of AI in its initial announcement of investments in four AI-driven companies. The first, Anthropic, bills itself as an “AI safety and research company,” trying to create more predictable and steerable AI systems, without the unintended consequences and bad behavior of some large AIs.

The second, Cohere, is a natural-language processing startup which, it said, aims to make language AI accessible to a much broader range of companies. Hearth AI, an agent-relationship management company, is dedicated to creating AI that can understand human social networks. Finally, You.com is an AI-powered search engine, offering summarized search results and a degree of privacy compared to large commercial offerings.

Salesforce said that it already uses AI technology for sales, service, marketing and commerce applications, which allows users to quickly analyze behavior and improve customer experiences in those areas.

Generative AI – that is, AI that can create its own information, whether that’s text, images or otherwise – is very much the technology of the moment, with ChatGPT and generative art program DALL-E 2 wowing users. Yet, Salesforce warns, there are real downsides to slapdash or careless development of generative AI systems. These can include learned toxicity, the possibility of widespread misuse (e.g. people passing off AI-generated content as their own), and more.

“As businesses race to bring this technology to market, it’s critical that we do so inclusively and intentionally,” said Kathy Baxter, principal architect for Salesforce’s ethical AI practice, and Paula Goodman, chief ethical and humane use officer, in a blog post Tuesday. “It’s not enough to deliver the technological capabilities of generative AI, we must prioritize responsible innovation to help guide how this transformative technology can and should be used.”

Salesforce’s existing AI offerings are grouped under the Einstein product family. Einstein technology currently offers predictive analytics, and today Salesforce announced that it is testing new software, Einstein GPT, that will offer generative AI.

Salesforce’s AI development has been guided, in large part, by a set of rules that echo the Rome Call for AI Ethics, a declaration that took place at the Vatican in 2020, where tech CEOs, religious leaders and representatives from government and academia pledged to develop AI that is unbiased, transparent, safe and inclusive.

A Salesforce survey released yesterday showed that, despite some skepticism, over half of businesses believe that “generative AI is a ‘game-changer,’” with strong majorities seeing the possibilities for reducing team workloads, selling products more effectively and leveraging data.

Artificial Intelligence, Enterprise Applications

All Posts

The US National Aeronautics and Space Administration (NASA) has overspent about $15 million on Oracle software over the past five years because it lacked a centralized software asset management practice, according to an audit report published by the space agency’s office of the inspector general (OIG).  

The report attributes the huge over-expenditure to vendor lock-in and NASA’s unwillingness to risk a license audit by Oracle because of its lack of visibility into software management.

Vendor lock-in, according to the report, is a situation when an enterprise customer using a product or service cannot easily transition to a rival product or service.

“NASA purchased large amounts of Oracle products to support Space Shuttle processing and other mission operations during that timeframe containing licensing terms that made transitioning to a competitor difficult due to proprietary technologies,” the OIG wrote in the report.

NASA was unwilling to commit to an Oracle audit as it was scared that the resultant penalties from the audit would cost more than the $15 million, the report showed.

“OCIO (office of the chief information officer) officials explained that they ‘knew better than to try our luck with an audit.’ Simply put, merely the potential threat of being audited by the vendor encouraged overbuying when the accuracy of agency software asset management was suspect,” the report said.

An email sent to Oracle about easing “lock-in” practices didn’t immediately receive a response.

Non-existence of a software asset management (EAM) program

The space agency’s problem, according to the report, is the absence of a centralized software asset management practice and its current “ad-hoc” practices, which could expose NASA to operational, financial, and cybersecurity risks.

Software asset management is the practice of controlling and optimizing the purchase, deployment, maintenance, and utilization of software applications or suites in an organization or institution.

“Efforts to implement an enterprise-wide software asset management program have been hindered by both budget and staffing issues and the complexity and volume of the agency’s software licensing agreements,” the OIG wrote in the report, giving the agency’s software management practices a “basic” rating—the lowest rating as per the International Organization for Standardization.

The agency uses over 49,000 desktops, laptops and engineering computers.

Further, the report showed that NASA was years away from moving to an enterprise computing model and was in violation of the federal policy to implement a centralized software asset management program that tracks inventory and license data.

“We also found internally developed mission and institutional software applications suffer from a lack of centralization and inventory visibility, limiting the agency’s ability to identify duplicative or obsolete software,” the OIG wrote.

In addition, NASA’s current organizational setup, which is against federal policy, hinders the effective implementation of a centralized software management policy.

“The agency’s software asset management office and software manager positions are misaligned and do not report to the chief information officer as required by federal policy,” the OIG wrote as part of the report.

Other challenges plaguing the space agency includes inconsistent processes for legal representation during software contract negotiations or vendor audits, unsupervised training software and unsupervised software buying.

These challenges expose the agency to increased costs because of penalties for violations of software licensing agreements, the report showed.

“NASA has failed to implement processes necessary to manage financial risks as software purchases are not sufficiently tracked and authorized by the Office of the Chief Information Officer (OCIO)—allowing some users to bypass OCIO authorization (and software asset management team scrutiny) to purchase software through alternative means such

as purchase cards,” the OIG wrote.

NASA overspent more than $35 million

The OIG also pointed out an additional $20 million expense in fines and overpayments, which could have been avoided.

“We estimate the agency could have saved approximately $35 million ($20 million in fines and overpayments and $15 million in unused licenses) and moving forward could save $4 million over the next 3 years by implementing an enterprise-wide software asset management program,” the OIG report said.  

According to the OIG’s analysis, almost 11,000 users, between 2020 and 2022, were granted privileged access (the ability to control one’s computer system akin to administrative rights) to download software at will due to operational constraints and delay in funding.

In 2017, NASA had to pay $18.9 million to IBM post an audit to bring its software usage in compliance with license agreements.

In 2021, multiple vendors such as SAP, Dassault and Ansys, collectively were paid about $4.4 million by the agency to settle software usage penalties.

NASA

ENDS

Asset Management Software, IBM, Oracle, SAP, SUSE