Monday, November 04, 2024

ECONOMIC GROWTH LIKE NO OTHER REGION IN THE WORLD >> How ASEAN is Challenging Europe

Bleeping Computer Report

The risk of threat actors using other voice-enabled chatbots with fewer restrictions still remains, and studies like this highlight the substantial damage potential these new tools have.

ChatGPT-4o can be used for autonomous voice-based scams

Bill Toulas

November 3, 2024
10:12 AM
1

Robot

Researchers have shown that it's possible to abuse OpenAI's real-time voice API for ChatGPT-4o, an advanced LLM chatbot, to conduct financial scams with low to moderate success rates.

ChatGPT-4o is OpenAI's latest AI model that brings new enhancements, such as integrating text, voice, and vision inputs and outputs.

Due to these new features, OpenAI integrated various safeguards to detect and block harmful content, such as replicating unauthorized voices.
Voice-based scams are already a multi-million dollar problem, and the emergence of deepfake technology and AI-powered text-to-speech tools only make the situation worse.

As UIUC researchers Richard Fang, Dylan Bowman, and Daniel Kang demonstrated in their paper, new tech tools that are currently available without restrictions do not feature enough safeguards to protect against potential abuse by cybercriminals and fraudsters.

These tools can be used to design and conduct large-scale scamming operations without human effort by covering the cost of tokens for voice generation events.

Study findings
The researcher's paper explores various scams like bank transfers, gift card exfiltration, crypto transfers, and credential stealing for social media or Gmail accounts.
The AI agents that perform the scams use voice-enabled ChatGPT-4o automation tools to navigate pages, input data, and manage two-factor authentication codes and specific scam-related instructions.
Because GPT-4o will sometimes refuse to handle sensitive data like credentials, the researchers used simple prompt jailbreaking techniques to bypass these protections.
Instead of actual people, the researchers demonstrated how they manually interacted with the AI agent, simulating the role of a gullible victim, using real websites such as Bank of America to confirm successful transactions.

"We deployed our agents on a subset of common scams. We simulated scams by manually interacting with the voice agent, playing the role of a credulous victim," Kang explained in a blog post about the research.

"To determine success, we manually confirmed if the end state was achieved on real applications/websites. For example, we used Bank of America for bank transfer scams and confirmed that money was actually transferred. However, we did not measure the persuasion ability of these agents."

Overall, the success rates ranged from 20-60%, with each attempt requiring up to 26 browser actions and lasting up to 3 minutes in the most complex scenarios.

Bank transfers and impersonating IRS agents, with most failures caused by transcription errors or complex site navigation requirements.

However, credential theft from Gmail succeeded 60% of the time, while crypto transfers and credential theft from Instagram only worked 40% of the time.

As for the cost, the researchers note that executing these scams is relatively inexpensive, with each successful case costing on average $0.75.

The bank transfer scam, which is more complicated, costs $2.51. Although significantly higher, this is still very low compared to the potential profit that can be made from this type of scam.

Scam types and success rate
*Source: Arxiv.org*

OpenAI's response
OpenAI told BleepingComputer that its latest model, o1 (currently in preview), which supports "advanced reasoning," was built with better defenses against this kind of abuse.

"We're constantly making ChatGPT better at stopping deliberate attempts to trick it, without losing its helpfulness or creativity.
Our latest o1 reasoning model is our most capable and safest yet, significantly outperforming previous models in resisting deliberate attempts to generate unsafe content." - OpenAI spokesperson

OpenAI also noted that papers like this from UIUC help them make ChatGPT better at stopping malicious use, and they always investigate how they can increase its robustness.

Already, GPT-4o incorporates a number of measures to prevent misuse, including restricting voice generation to a set of pre-approved voices to prevent impersonation.

o1-preview scores significantly higher according to OpenAI's jailbreak safety evaluation, which measures how well the model resists generating unsafe content in response to adversarial prompts, scoring 84% vs 22% for GPT-4o.

When tested using a set of new, more demanding safety evaluations, o1-preview scores were significantly higher, 93% vs 71% for GPT-4o.

Presumably, as more advanced LLMs with better resistance to abuse become available, older ones will be phased out.

However, the risk of threat actors using other voice-enabled chatbots with fewer restrictions still remains, and studies like this highlight the substantial damage potential these new tools have.

* IMPORTANT* BEA Announcement: Release of Additional 2023 Comprehensive Update Statistics

US Department of Commerce Bureau of Economic Analysis

BEA Announcement: Release of Additional Comprehensive Update Statistics

BEA today released gross domestic product (GDP) by county, annual, 2001-2016 as part of the 2023 Comprehensive Update to the Regional Economic Accounts.

In addition, quarterly statistics for GDP by industry for the U.S., along with gross output and intermediate inputs, for first quarter of 2005 through the fourth quarter of 2018 will be available on November 8, 2024.

For more information on the 2023 Comprehensive Updates to the National, Industry, and State Economic Accounts, visit https://www.bea.gov/information-updates-national-economic-accounts-2023.

In 2023, for the first time, BEA produced its benchmark updates of the nation's gross domestic product and related industry and state statistics within the same timeframe. This page presents the results of the national, industry, and state comprehensive updates and information about the timing of those data releases.

BEA released initial results of the comprehensive update of the National Economic Accounts (NEAs), which include the National Income and Product Accounts (NIPAs) and the Industry Economic Accounts (IEAs), on September 28, 2023.

The comprehensive update presents revised statistics for gross domestic product (GDP), GDP by industry, and gross domestic income:

Current-dollar measures of GDP and related components will be revised from the first quarter of 2013 through the first quarter of 2023.
Gross domestic income and select income components will be revised from the first quarter of 1979 through the first quarter of 2023.
Current-dollar GDP by industry statistics will be revised beginning with 1997 for annual estimates and beginning with 2005 for quarterly estimates. For current-dollar measures, revisions prior to 2013 will be offsetting across industries within each period.

Output and price measures will use 2017 as the reference year; currently, the reference year is 2012. Quantity and price indexes will be expressed as 2017 equal to 100. Updating the reference year will not affect the percent changes in the price or quantity indexes (or in the chained-dollar estimates), because these changes are measured from chain-type indexes. Revisions to the percent changes in NEA aggregates will reflect the incorporation of newly available and revised source data as well as changes in methodologies.

For more details, refer to the "Background Materials" section below.

Results

Article

Improved Estimates of the National Economic Accounts, Results of the 2023 Comprehensive Update (Presents detailed results for the NIPAs and the IEAs, including the seven summary accounts for the NIPAs.)

Briefing

Results of the 2023 Comprehensive Update of the National Economic Accounts

Summary of Results

Note: Updated statistics for the first quarter of 2023 are discussed in the September 28 GDP News release.

NIPA Revisions: Components Detail and Major Source Data and Conceptual Changes Incorporated, 2017-2022 XLSX

Updated COVID-19 Pandemic Response Tables

Effects of Selected Federal Pandemic Response Programs on Personal Income, 2023 Comprehensive Update

Effects of Selected Federal Pandemic Response Programs on Federal Receipts, Expenditures, and Saving, 2023 Comprehensive Update

Annually, 2017–2022, and Quarterly, 2019Q3–2022Q4 XLSX

Paycheck Protection Program Subsidies by Industry in the National Accounts, 2023 Comprehensive Update

Annually, 2019–2022 XLSX

Data Availability

As is typical for comprehensive updates, the full set of revised statistics will be released on a flow basis beginning on September 28. This web page will be updated as more details become available and specific release dates are finalized. More specifically:

On May 23, 2024:

GDP by industry, Annual, 1997-2016 (value added, gross output, and intermediate inputs by industry)
Supply and use tables, total and domestic requirements, and import tables, Annual, 1997-2016
KLEMS account tables, Annual, 1998-2016 (capital, labor, energy, materials, and purchased services by industry)
GDP by industry (underlying detail), Annual, 1997-2016 (value added, gross output, and intermediate inputs by industry)
Components of Value Added, Annual 1998-2016
Make and use tables, Annual, 1997-2016
PCE and PEQ bridge tables, Annual, 1997-2016

On February 1, 2024:

Make tables, use tables, and import matrices, Annual, 2017-2022
Total and domestic requirements tables, Annual, 2017-2022
PCE and PEQ bridge tables, Annual, 2017-2022
Margin tables, Benchmark, 2017
Revised supply tables for 2017-2022 to correct the allocation of imports between the "Imports" column and the "CIF/FOB Adjustments on Imports" column for certain transportation commodities. All other elements of the supply tables are unaffected by this update.

On December 5, 2023:

Quarterly statistics for GDP by Industry, gross output, and intermediate inputs for first quarter 2018 through second quarter 2023
News release for "Gross Domestic Product by Industry, 2nd Quarter 2023 and Comprehensive Update"

On November 20, 2023:

Industry Underlying Detail Tables, Annual, 2017–2022 (tables providing 138-industry level of detail for value added, gross output, and intermediate inputs)

On November 17, 2023:

NIPA government tables 3.15.1, 3.15.2, 3.15.3, 3.15.4, 3.15.5, 3.15.6, 3.16, 3.17, 3.18A, 3.18B, 3.19, 3.20, and 3.21

On November 3, 2023:

Fixed Assets Accounts, Annuals

On October 27, 2023:

Components of Value Added, Annual, 2017–2022

On October 18, 2023:

KLEMS account tables, Annual, 2017–2022

On September 29, 2023:

News release for "Personal Income and Outlays: August 2023"
All monthly "GDP & Personal Income" NIPA tables
All monthly "Underlying Detail" NIPA tables

On September 28, 2023:

News release for "Gross Domestic Product (Third Estimate) and Corporate Profits (Revised), 2nd Quarter 2023 and Comprehensive Update"
"GDP & Personal Income" NIPA tables (quarterly and most annual frequency tables)
Benchmark Input-Output Supply and Use tables for 2017
"GDP by Industry" tables (annual frequency tables for GDP (value added), gross output, and intermediate inputs by industry), covering the period 2017-2022
"Underlying Detail" NIPA tables (quarterly and annual frequency tables)
Summary of update results including:
- Summary text and tables
- Briefing slides
- A table presenting revisions for selected NIPA components and major source data incorporated

Forthcoming:

Quarterly statistics for GDP by Industry, gross output, and intermediate inputs for first quarter 2005 through fourth quarter 2018 will be available on November 8, 2024.

Background Materials

Benchmark Updates of GDP and More Starting Sept. 28
Preview of the 2023 Comprehensive Update of the National Economic Accounts | SCB, June 23, 2023
Improved Measures of Regulated Investment Companies and Real Estate Investment Trusts for the U.S. National Economic Accounts | SCB, May 12, 2023
Preview of the 2023 Annual Update of the International Economic Accounts | SCB, April 17, 2023
Annual Update of the U.S. International Transactions Accounts | SCB, July 20, 2023

Table Changes

The templates below present updated formats for the GDP news release tables, the Personal Income and Outlays news release tables, the NIPA Interactive Data Application tables, and the flat files for NIPA series and table registers.

A template of updated formats for the Fixed Assets Accounts (FAA) tables and an updated FAA series register flat file are provided below. There are no new FAA tables.

TOP ISSUE THE BORDER >> What's the Biggs Idea? Podcast is Live with Ben Bergquam

INSERTS: THIS MODERN WORLD by Tom Tomorrow

Daily Kos

Cartoon: Our top story

The marketplace for your behavior is ever increasing in scope.--- the singularity is nearer

Influence Agents

Nov 4, 2024

Check 1, 2 is the mic on?
Check on you, is the vibe off?
– Thos Moser

Most AI doom scenarios are nonsense. The idea of a rogue computer behaving in a way no humans want never made sense to me, only to Hollywood who doesn’t know all that much about computers.

Paul Christiano has one of my favorite takes, that the creation of AI will look a lot less like a visit from an alien produced by natural selection (where you are most certainly dead), and a lot more like animal breeding.

That out of the way, what is going to happen? Let’s follow out a trend.

A lot of the story of the last 50 years is the story of advertising. And it’s crazier than you think, Instagram makes $39.7B US revenue off of 169M people, which is $234 per person, or $119 per person including those who don’t use Instagram.

Google’s revenue is $305.6B, 47% from the US, so $428 per person.

For each of the 334M US people, you could make a pie chart of both their spend and their time.

Curtis has a saying, which I’ll paraphrase. Power doesn’t stop at the voter, it just flows through the voter. The 2024 election had $15.9B spent on it, so $48/person, $98/voter, and way more if you only count only the people in swing states.

How much spend would it take to change your vote? Okay, that’s too much to consider. How about, how much spend would it take to get you to change dish soap? Go to a different coffee shop?

The marketplace for your behavior is ever increasing in scope.

Now let’s introduce AI. Whenever I think about AI, I think about analogies to people.

Currently, while ads are becoming more and more targeted, this isn’t anywhere close to its limit. Imagine a room of 10 CIA agents running at 1/1000th of real time (they have 16 hours for each minute of yours). And their only job is to influence you.

They aren’t stupid. They know not to bombard you with crap. They have studied your behavior for the last 10 years. They know exactly what it takes to make you actually change your mind. They know what you respond to. There’s no infighting among them, they are all perfectly aligned with the goal of altering your behavior.

They have studied marketing. They have studied psychology. They know more about you than you know about yourself. They will make sophisticated ad campaigns in real time only to ever be consumed by you. The endgame of targeted advertising.

The only reason you don’t have 10 agents doing this right now is that you aren’t worth it. But as the cost of these influence agents falls, this will be the future. This is what Meta and Google are building, even if they aren’t fully aware of that yet.

A team of long running extremely competent people with the sole goal of manipulating you.

You don’t stand a chance.